The effects of sleep restriction on NREM stage 2 sleep in Panel A; on NREM slow wave sleep (SWS) in Panel B; and on REM sleep in Panel C. Data are adapted from Van Dongen et al.12 Following 8 hours of time in bed on baseline nights (B1, B2, B3), sleep was restricted for 14 consecutive nights to either 4 hours of time in bed (•, n = 13 healthy adults), 6 hours of time in bed (▲, n = 13), or 8 hours of time in bed (■, n = 9). Restriction was implemented by delaying bed time and holding sleep offset time constant (07:30). Sleep restriction nights were followed by 3 nights of 10 hours of time in bed for recovery sleep (R1, R2, R3). Sleep stages were scored polysomnographically for 2 out of every 3 nights during the experiment. Panel A: During the 14 nights of restriction to 4 h of time in bed, NREM stage 2 sleep was decreased an average of more than 2 h per night relative to the 8-h control condition (p < 0.001). Stage 2 sleep was decreased approximately 1 h per night in the 6-h condition relative to the control condition (p < 0.001). Panel B: In contrast to NREM stage 2 sleep, NREM slow wave sleep (SWS) showed no significant reduction in either the 4-h or 6-h sleep restriction conditions relative to the 8-h control condition. Panel C: Relative to the 8-h control condition, REM sleep was reduced by approximately 47 minutes a night during the 14 nights of restriction to 4 h time in bed (p < 0.01), and by 24 minutes a night during the 14 nights of restriction to 6 h time in bed (p < 0.05).

Digit symbol substitution task (DSST) performance responses to varying doses of daily sleep across 14 days. Data from n = 35 subjects (8h condition n = 9, 6h condition n = 13 and 4h condition n = 13). Mean DSST per day (07:30–23:30), measured at 2-h intervals expressed relative to baseline (BL). The curves represent statistical nonlinear model-based best-fitting profiles of the DSST performance response to sleep loss. Adapted from Van Dongen et al.12

Mean (SEM) plasma high-sensitivity C-reactive protein (CRP) in n = 4 subjects undergoing 10 consecutive nights of sleep restricted to 4.2 h time in bed, and in n = 5 control subjects who had 10 consecutive nights of sleep restricted to 8.2 h time in bed (closed squares). Significance of difference in change from baseline to day 10 between groups (p = 0.08 for interaction) by mixed-models analysis of variance on log-transformed data: the change from baseline to day 10 for the 4-h sleep restriction group was significant (p = 0.05), whereas the change from baseline to day 10 in the 8-h control group was not (p = 0.72). Figure adapted from Meier-Ewert et al.71

The effects of varying doses of nocturnal sleep time on lapses of attention from the psychomotor vigilance test (PVT). Panel A from Van Dongen et al. 12 involved experimental sleep restriction of n = 36 healthy adults for 14 consecutive nights. In this experiment sleep was restricted for 14 consecutive nights. Subjects were randomized to 4 h time in bed (n = 13), 6 h time in bed (n = 13), or 8 h time in bed (n = 9). PVT performance was assessed every 2 h (9 times each day) from 07:30 to 23:30. The graph shows systematic increases in lapses of sustained attention when sleep was restricted to either 4 h (p < 0.001) or 6 h (p < 0.001) per night, but not when sleep was restricted to 8 h per night (p = 0.29). The increase in lapsing was worse in the 4-h sleep condition than in the 6-h sleep condition (p = 0.036), further supporting a dose-response relationship within and between conditions. The horizontal dotted line shows the level of lapsing found in a separate experiment when subjects had been awake continuously for 64–88 h. For example, by day 7, subjects in the 6-h sleep restriction condition averaged 54 lapses (6 lapses × 9 test times) that day, while those in the 4-h sleep condition averaged 70 lapses that day. Panel B shows comparable sleep restriction data from Belenky et al.10 In this study sleep was restricted for 7 consecutive nights in n = 66 healthy adults. They were randomized to 3 h time in bed (n = 13), 5 h time in bed (n = 13), 7 h time in bed (n = 13), or 9 h time in bed (n = 16). Performance was assessed 4 times each day from 09:00 to 21:00. PVT lapses increases steadily across days in the 3-h (p = 0.001) and 5-h (p = 0.001) sleep restriction conditions (PVT response speed, but not lapses, was reduced in the 7-h condition, not shown). As in Panel A, the horizontal dotted line shows the level of lapsing found in a separate experiment when subjects had been awake continuously for 64–88 h.12 Considering data in both Panels A and B, it is clear that restriction of nocturnal sleep time to <7 h per night in healthy adults results in systematic increases in lapses of waking attention that get progressively worse across days, in a dose-response manner.

Data from n = 35 subjects (8h condition n = 9, 6h condition n = 13 and 4h condition n = 13). Restriction of nocturnal sleep in healthy adults resulted in near-linear increases in Psychomotor Vigilance Test (PVT) lapses of attention across 14 days (coefficients of change near 1.0), but subjective ratings of sleepiness and fatigue (regardless of the psychometric scale used) showed a nonlinear coefficient below 0.5 for change over days. This meant that as objective performance continued to decline near-linearly, there were only minor further increases in the subjective ratings of sleepiness. By the end of the 14 days of sleep restriction, when performance was at its worst levels, subjects in the 4-h and 6-h sleep period conditions reported feeling only slightly sleepy. Therefore, unlike performance measures, sleepiness ratings appeared to show adaptation to chronic partial sleep deprivation. The lack of reports of intense feelings of sleepiness during chronic sleep restriction may explain why sleep restriction is widely practiced—people have the subjective impression they have adapted to it because they do not feel particularly sleepy. Adapted from Van Dongen et al.12