Electroencephalogram (EEG) shows long-range power-law and multifractal scaling behaviors, which varies with physiological and pathological conditions. In the present paper, the scaling behavior of sleep EEG is studied using monofractal detrended fluctuation analysis (DFA) and multifractal singularity spectrum. It is found that, the DFA scaling exponent α is more sensitive to the differences between EEG derivations and subject groups, but irregular with sleep stages. However, the variation of multifractal singularity strength range Δα with sleep stages is more regular, it increases constantly from sleep stage Ⅰ to stage Ⅳ, showing less difference between EEG derivations and subject groups. Thus, it is suggested that multifractal parameter Δα is more suitable than monofractal scaling exponent α to be used as a quantitative parameter for sleep staging.