PURPOSE: To assess the physiological meanings of the detrended fluctuation analysis (DFA) slope alpha and its relationship to spectral measures in heart rate variability, this study investigated changes of alpha and its corresponding spectral measures over various night-sleep stages. METHODS: The overall DFA alpha and natural-logarithm-transformed power values of the spectral parameters ln[high-frequency (HF)], ln[low-frequency (LF)], and ln[very-low-frequency (VLF)], and their relationship from one 5-min proper electrocardiography segment in each of pre-sleep-wakefulness (AWK), non-rapid eye movement stage 2 (N2), slow-wave (N3), the first and the latest rapid-eye movement sleep (REM1, and REM2), were computed in 93 otherwise healthy males (44.1 +/- 7.7 years.) with wide-ranged apnea-hypopnea, periodic-limb movement and arousal indices (19.0 +/- 20.9, 4.7 +/- 9.9, and 10.7 +/- 18.2 h, respectively). RESULTS: While ln(HF) dipped from AWK, N2, and N3 to REM1 then rebounded to the origin level at REM2, ln(VLF) dipped from AWK to N2, N3 trough, and then surged to levels surpassing AWKs and N2s at REM1 and REM2. ln(LF/HF), ln(VLF/HF), and alpha dipped from AWK and N2 to N3 trough, surged to levels surpassing AWKs, and N2s at REM1 then became attenuated at REM2. By general linear modeling, the relationship between alpha and the corresponding spectral values can be seen over various stages as alpha = b 0 + 0.147 x ln(VLF/HF) (R (2) = 0.766), regardless of age and sleep-sympathoexcitatory episodes. CONCLUSION: The REM sleep attenuations appeared in ln(HF) and its derivatives, such as ln(LF/HF), ln(VLF/HF), and the overall DFA slope alpha values. The quantitative function of ln(VLF/HF) describes the alpha values constantly for overnight sleep stages, and it is not affected by age, LF, PLM, and AHI. Our findings therefore suggest that in sleep studies with spectral HRV measures, ln(VLF/HF) as a surrogate of the overall DFA slope alpha should be calculated at the same time.