In the diagnostic work-up of hypersomnias of central origin, the complaint of excessive daytime sleepiness should be objectively confirmed by MSLT findings. Indeed, the features and diagnostic utility of spontaneous daytime sleep at 24 h continuous polysomnography (PSG) have never been investigated. We compared daytime PSG features to MSLT data in 98 consecutive patients presenting with excessive daytime sleepiness and with a final diagnosis of narcolepsy with cataplexy/hypocretin deficiency (n = 39), narcolepsy without cataplexy (n = 7), idiopathic hypersomnia without long sleep time (n = 19), and ‘hypersomnia’ with normal sleep latency at MSLT (n = 33). Daytime sleep time was significantly higher in narcolepsy-cataplexy but similar in the other groups. Receiver operating characteristics (ROC) curves showed that the number of naps during daytime PSG predicted a mean sleep latency ≤8 min at MSLT with an area under the curve of 0.67 ± 0.05 (P = 0.005). The number of daytime sleep-onset REM periods (SOREMPs) in spontaneous naps strikingly predicted the scheduled occurrence of two or more SOREMPs at MSLT, with an area under the ROC curve of 0.93 ± 0.03 (P < 10−12). One spontaneous SOREMP during daytime had a sensitivity of 96% with specificity of 74%, whereas two SOREMPs had a sensitivity of 75%, with a specificity of 95% for a pathological REM sleep propensity at MSLT. The features of spontaneous daytime sleep well correlated with MSLT findings. Notably, the occurrence of multiple spontaneous SOREMPs during daytime clearly identified patients with narcolepsy, as well as during the MSLT.