As a first step in a program designed to study the central control of the heart rate variability (HRV) during sleep, we conducted polysomnographic and electrocardiogram recordings on chronically-prepared cats during semi- restricted conditions. We found that the tachogram, i.e. the pattern of heart beat intervals (RR intervals) was deeply modified on passing from alert wakefulness through quiet wakefulness (QW) to sleep. While the tachogram showed a rhythmical pattern coupled with respiratory activity during non-REM sleep (NREM), it turned chaotic during REM sleep. Statistical analyses of the RR intervals showed that the mean duration increased during sleep. HRV measured by the standard deviation of normal RR intervals (SDNN) and by the square root of the mean squared difference of successive intervals (rMSSD) were larger during REM and NREM sleep than during QW. SD-1 (a marker of short- term variability) and SD-2 (a marker of long-term variability) measured by means of Poincare plots increased during both REM and NREM sleep compared to QW. Furthermore, in the spectral analysis of RR intervals, the band of high frequency (HF) was larger in NREM and REM sleep in comparison to QW, whereas the band of low frequency (LF) was larger only during REM sleep in comparison to QW. The LF/HF ratio was larger during QW compared either with REM or NREM sleep. Finally, sample entropy analysis used as a measure of complexity, was higher during NREM in comparison to REM sleep. In conclusion, HRV parameters, including complexity, are deeply modified across behavioral states.