They used auditory and mechanical stimuli to elicit EEG microarousals with a frequency of ≥ 30 events/h. If the stimuli failed, the subsequent stimulus got larger by increasing the tone volume of auditory stimuli or combining the two kinds of stimulus. In this study, most of the stimuli were effective, including the last one at the end of the sleep period, which was probably the largest stimulus in intensity. When a human is exposed to noxious or potentially noxious stimuli, there is an increased secretion of corticotropin and, consequently, a rise in the circulating cortisol. Approximately 90% to 95% of the cortisol in the plasma binds to plasma proteins, which slows the elimination of cortisol from the plasma. Therefore, cortisol has a relatively long half-life of 60 to 90 min and thus a lasting action.5 An increased morning cortisol level (measured at 8:00 am in this study) may be the consequence of the last effective stimulus, elicited 1 or 2 h prior to measurement, rather than that of sleep fragmentation. As the authors stated, elevations of cortisol, even within the normal physiologic range, can decrease insulin sensitivity, enhance hepatic gluconeogenesis, and inhibit insulin secretion.6 To settle this dispute, one additional night of nonfragmented sleep following the two nights of fragmented sleep is needed as a sleep-recovery period, as stated in some sleep deprivation studies.7-9 In addition, one single stimulus is given near the end of the night, which is of the same intensity and timing as the last stimulus in the previous night of sleep fragmentation. If an elevated cortisol level is still observed in the coming morning (day 5), the disturbed glucose metabolism may not be attributed to the effect of sleep fragmentation.