Abstract: To explore electroencephalogram (EEG) oscillation playing a key role during sleep transitions, detrended fluctuation analysis (DFA) was used to calculate the scaling exponents of various EEG oscillations during different vigilance states. The coefficients of variations of these scaling exponents between or among different vigilance states are calculated. The results reveal that the scaling exponents among the three sleep stages are significantly different for each oscillation; low-frequency EEG oscillations possesses long-range temporal correlations, and the coefficients of variations of their scaling exponents change vastly during sleep transitions. Thus sleep alternations may be achieved by the processes similar to avalanches, and low-frequency EEG oscillations may play a key role in sleep transitions.