Recent developments in the understanding of the neurophysiological mechanisms underlying electroencephalographic (EEG) slow waves and sleep spindles imply an inverse relationship between these two EEG activities. The interrelationship between slow wave activity (0.75-4.5 Hz) and sleep-spindle activity (12-15 Hz) in electroencephalograms recorded in nine male subjects, during nocturnal baseline sleep and during recovery sleep from 40 h of wakefulness, was analyzed by power spectral analysis based on the fast Fourier transform and by transient patterns detection algorithms. Both techniques revealed that spindle activity was highest in sleep stage 2, increased over consecutive non-rapid-eye-movement sleep (non-REM sleep) episodes and was suppressed during recovery sleep. In contrast, slow wave activity decreased over consecutive non-REM sleep episodes and was enhanced during recovery sleep. Analysis of the dynamics of spindle and slow-wave activity within non-REM sleep episodes demonstrated that in the initial 20% of these episodes both spindle activity and slow wave activity increased, whereafter slow wave activity continued to increase but the average amplitude of spindles and total spindle activity, but not spindle density, decreased. At the end of non-REM sleep episodes the reverse pattern was observed. Sleep deprivation induced a more rapid rise of both spindle and slow wave activity in the very beginning of sleep. These data demonstrate that when averaged per sleep episode or non-REM sleep episode an inverse relationship between SWA and spindle activity exists but that in the initial and final part of non-REM sleep episodes the association between these two activities is positive. This biphasic relationship is discussed with reference to the hypothesis that the transition from sleep spindles to slow waves is dependent on a progressive hyperpolarization of thalamo-cortical neurons.