Microstructural electroencephalographic changes during the wake fulness-sleep transition have been investigated by comparing two definitions of sleep onset: the first occurrence of stage 1 and of stage 2. Power values were calculated across a 1-28-Hz frequency range in a 1-Hz bin resolution in the sleep recordings of 26 normal subjects. Quantitative changes were assessed after averaging individual time series, aligned with respect to the first occurrence of stage 1 or of stage 2. The time course of the single-Hz activity revealed a linear increase of power in the 1-6-Hz range and a linear decrease in the 9-12- and 16-28-Hz ranges during the stage 1 transition. During the stage 2 transition, electroencephalogram power linearly increased in the 1-7- and 14-15-Hz ranges and decreased in the 18-28-Hz range, while the 8-12-Hz range fitted a second-order polynomial curve. The two 'switch' points were also compared in their ability to differentiate Hz by Hz wakefulness from sleep: a lower mean power was found after stage 1 onset in the 9-11-Hz and 20-28-Hz bins and a higher one in the 1-5-Hz bins, while a higher power was found in the 1-8-Hz and 12-16-Hz bins and a lower one in 18-28-Hz bins after stage 2 onset. The time course of three electroencephalographic frequency ranges [delta/theta/sigma (1-7 and 12-16 Hz); beta (17-28 Hz); alpha (8-11 Hz)], grouped on the basis of a principal component analysis, fitted a first-order polynomial curve for the first two ranges, and a second-order polynomial curve for the last. with a progressive decrease curing wakefulness, a minimum point during stage 1, and a subsequent increase during stage 2. The uniformly increasing electroencephalographic power across the 1-16-Hz frequency range during stage 2 and the shift of functional meaning for the alpha power during stage 1 point to the start of stage 2 as a more reliable boundary between wakefulness and sleep. (C) 2001 IBRO. Published by Elsevier Science Ltd. All rights reserved.