REM-NREM cycle in the cat may be sleep-dependent.

The periodicity of the rapid eye movement-nonrapid eye movement (REM-NREM) cycle in real time versus compressed sleep was determined by autocorrelation, computed on the sequence of sleep stages in recordings from spontaneously sleeping cats. The resulting autocorrelation function was correlated to damped cosine waves, and the highest squared correlation coefficient (r2) was taken as indicating the most likely periodicity in the data entered for each animal. The periodicity of REM sleep was stronger (significantly higher r2) in the compressed sleep data than in the real-time data, indicating sleep dependency of the REM-NREM cycle. The REM-NREM cycle lengths determined by the autocorrelation technique were not significantly different for the real-time and compressed sleep data. The REM sleep episode interval, defined as the average interval between the start of successive REM sleep episodes, was significantly shorter for real-time sustained sleep than the cycle lengths as determined by the autocorrelation technique. A model is proposed which explains this phenomenon as due to fragmentation of REM sleep within the time periods with high probability for REM sleep. When such fragmentation occurs, the average REM sleep episode interval will not reflect an ultradian REM sleep periodicity.