It is commonly stated that vigorous exercise close to bedtime impairs sleep (1,2,20), perhaps because of physiological arousal. However, this assumption has not been consistently confirmed by surveys (19,33,35) or by experimental evidence (16,28).
In a random sample, 1190 adults in Tampere, Finland (33) were asked to identify the most important factors disturbing sleep. Late night exercise was mentioned by less than 8% of the men (N = 430) and less than 6% of the women (N = 555). Of those who reported exercising habitually late in the evening (i.e., after 8 p.m.) (N = 320), the majority indicated that late evening exercise enhanced their ability to fall asleep (65%), increased the depth of sleep (62%), and resulted feeling better in the morning (60%). Although it was noted, "the frequency of negative impacts increased when the exercise was performed vigorously late at night" compared with earlier in the day, difficulty falling asleep, restless sleep, early awakening, and feeling more tired in the morning were reported by only 8-24% of the late-night exercisers. It was concluded that "a considerable proportion of vigorous late exercisers (after 8 p.m.) reported mainly positive effects (on sleep)" (19,35)
We have located only two experimental studies that have examined the influence of vigorous exercise within 4 h of bedtime on sleep. In a within-subjects, counterbalanced study of six males aged 23-28, Desjardins et al. (16) found no differences in sleep latency, movement time, stages 1, 2, 3, 4, or wakefulness after sleep onset (WASO) between a sedentary control condition and moderate or vigorous treadmill running completed 2 h before bedtime. In a within-subject, counterbalanced study of eight aerobically fit, good sleepers, O'Connor et al. (28) found no differences in a behavioral measure of sleep latency (i.e., response to auditory stimulation), wakefulness after sleep onset, or sleep efficiency, following 60 min of cycling at 60% of O2peak completed 30 min before bedtime compared with a sedentary control condition.
These studies suggest that exercise close to bedtime may not disturb sleep. Indeed, there are theoretical arguments that late-night exercise may be more likely to promote sleep than exercise at other times of day. First, exercise may promote sleep via anxiolytic effects, which are especially well documented during the first 2 h after moderate to vigorous exercise. Second, multiple lines of evidence indicate an interaction between central temperature regulation and sleep (25), so body cooling may be the "trigger" for sleep onset (27). Since temperature increases during exercise in proportion to exercise intensity, it is plausible that during recovery from vigorous late-night exercise, there may be a period of increased sleep propensity associated with brain cooling, which occurs for only a few hours. Exercise might be a welcome adjuvant or alternative treatment for insomnia because exercise is associated with improved mental and physical health and decreased mortality. In contrast, sleeping pills, the most commonly used insomnia treatment, are associated with tolerance (3), negative side effects (3), and increased mortality (23). Some behavioral treatments for poor sleep are expensive and require skilled clinicians. In light of an apparent trend of increasingly common night-time exercise in our society, there is a need to more clearly establish the sleep consequences of exercise close to bedtime.
The purpose of this experiment was to examine the influence of prolonged, exhaustive presleep exercise on sleep. The data were derived from a study examining the influence of exercise in combination with exposure to bright light on the circadian rhythm of melatonin excretion (37). Consequently, the exercise was not performed near the participants' usual bed time, but rather during the late-night or early morning (∼1000-0500 h). Thus, it is possible that sleep-impairing effects of exercise may have been overridden by homeostatic and circadian mechanisms. Nonetheless, an extreme exercise stimulus was thought to provide a strong test of whether exercise close to bedtime disrupts sleep.
Participants. Participants were 16 highly fit competitive male cyclists, ages 22-36. The group had a mean age of 27.3 yr (± SD 4.3), height 176.3 cm (± 6.3), weight 71.7 kg (± 8.1) and O2peak of 56.2 mL·kg−1·min−1 (± 2.9), respectively. Freedom from sleep disorders was verified by the Pittsburgh Sleep Quality Index (13) (global sleep disturbance index of less than 5). Participants had no recent shiftwork experience (previous year) or travel across multiple time zones (previous 4 wk). All participants signed informed consent and were paid to participate. The study was approved by the University of California, San Diego Human Research Institutional Review Board.
Baseline. Each subject's home sleep-wake schedule was reported on a daily sleep log for a baseline week. During the baseline, participants completed a graded cycle ergometry test to volitional exhaustion. Peak aerobic capacity (O2peak) was defined by a plateau of oxygen consumption with increasing work rate or by heart rate within 10 beats·min−1 of age-predicted maximum plus a respiratory exchange ratio greater than 1.10. The maximal heart rate obtained at O2peak was then used to establish the target heart rates.
Laboratory testing. Following the baseline period, each subject completed two 60-h laboratory protocols (3 nights and 2 d): a baseline night (NIGHT 1), followed by the experimental treatment (Exercise + Bright Light or Bright Light Alone; NIGHT 2), and a recovery night (NIGHT 3). The protocols were separated by 2-4 wk and counterbalanced in order.
Illumination was maintained at 3000 lux during the experimental treatments, 0 lux during the sleep periods, and 50 lux during all other times. Bright light was supplied by 1600 W cool-white fluorescent light bulbs built into the ceiling, providing 3000 lux illumination evenly throughout the room, although levels were lower or higher if the subjects glanced at the floor or ceiling, respectively. Ambient temperature was maintained at 20-21°C.
On NIGHT 1 and NIGHT 3, participants maintained their usual sleep-wake schedule established during baseline. Participants remained awake throughout their normal wake periods during DAY 1 and DAY 2, as verified by video monitoring.
In the Exercise + Bright Light condition (NIGHT 2), participants cycled at 65-75% of heart rate reserve for 3 h while exposed to 3000 lux light. Exercise intensity was verified by continuous heart rate recording with a Polar Heart Rate Monitor (Polar Electro, Inc., Port Washington, NY) which activated alarms when the subjects were out of their target zone. In the Bright Light Alone treatment (NIGHT 2), subjects were exposed to bright light of 3000 lux for 3 h while reading and/or watching TV in a seated position. Volunteers were asked to remain awake, as verified by continuous video recording. Timing of both 3-h treatments was centered at 6 h before the volunteer's usual wake time. Following completion of both treatments, participants took a 5-10 min shower. Bedtime was precisely 30 min after completion of the 3-h treatments. Participants were awakened at their normal time of awakening on the following morning (DAY 2). Thus, the duration of the sleep period on NIGHT 2 was only 4 h.
During all hours that were not devoted to sleep or to the experimental treatments, participants had access to food and nonalcoholic beverages. Moderate caffeine consumption (≤ one cup of coffee or 100 mg of caffeine from two cola drinks) was permitted during the first 4 h after arising. Coffee was prepared outside of the subject rooms by the staff. All other food and beverage was supplied by the staff; thus verification of the caffeine restriction was monitored. Participants spent the time in the laboratory at leisure, for example, studying or watching movies. Participants were free to walk about their rooms; however, vigorous exercise was prohibited outside of the experimental treatment.
An Actillume wrist monitor (Ambulatory Monitoring, Inc., Ardsley, NY) was worn throughout the stay in the laboratory. An algorithm for estimating sleep/wake from wrist movement using this device has been validated (15). The primary dependent variables were sleep onset latency (SOL), wakefulness after sleep onset (WASO), and total sleep time (TST). Subjective ratings of SOL (SSOL), WASO (SWASO), and of insomnia (SINSOMNIA) via a 100 visual-analog scale (range: 0 = No Insomnia to 100 = Worst Possible Insomnia) were assessed each morning 10 min after arising. Written instructions for the subjective ratings were as follows:
- SSOL: "Please estimate how many minutes it took you to fall asleep." Orally the volunteers were told that this was from the time in which the lights were turned out.
- SWASO: "Please estimate the minutes you were awake during the night after falling asleep." Orally the volunteers were told that this was the total time awake from the time they fell asleep until the time they arose in the morning.
- STST: "Please estimate your total sleep time during the night, not including time in which you were trying to fall asleep or time in which you were awake during the night (in other words, this should equal the time from when you fell asleep to the time of final awakening minus any time in which you were awake during the night."
- SINSOMNIA "Please rate the extent of your insomnia by drawing a vertical line through the line at the point that best describes your sleep (from no insomnia to worst possible insomnia." Orally volunteers were told that insomnia could mean trouble falling asleep and/or trouble with being awake during the night.
Rectal temperature (Tre) was measured each night from approximately 4 h before bedtime until wake time with a disposable probe (Yellow Springs, Instruments, Yellow Springs, OH; insertion depth 10 cm). The Actillume monitor stored temperature measurements every minute.
Analysis. Comparisons in sleep between Exercise + Bright Light and Bright Light Alone were analyzed for each night by paired t-tests. The NIGHT effects (e.g., the effect of delaying bedtime on NIGHT 2) on sleep were assessed by repeated measures ANOVA. The influence of the experimental stimuli on phase shifts in the circadian rhythm of melatonin excretion was also examined. Since circadian phase shifts can influence sleep (e.g., jet lag), Spearman rank-order correlations between circadian phase-shifts and the sleep variables were calculated. In addition, the sleep data for NIGHTS 2 and 3 were further analyzed by ANCOVA, using the circadian phase shifts elicited by the conditions as covariates. Temperature at bedtime, mean Tre during the sleep period, and the decline in Tre from bedtime to Tre minimum during sleep on the experimental night (NIGHT 2) were compared between conditions with paired t-tests. Spearman rank-order correlations examined the association between the temperature and sleep variables for NIGHT 2. The sleep data for NIGHT 2 were further assessed via ANCOVA, controlling for the temperature variables.
One subject stopped after 2 h of exercise (rather than 3 h) because of nausea. His data were excluded from the analyses. All other subjects completed both conditions without major difficulties and spent less than 10 min outside of the target heart rate zone.
Actillume-assessed sleep data were lost from one subject because of equipment failure. Descriptive data for SOL, WASO, and TST are displayed in Table 1. No significant differences in SOL, WASO, or TST were found between Exercise + Bright Light and Bright Light Alone on any of the nights (paired t-tests > 0.05). Significant NIGHT effects were found for SOL (F (2,12) = 8.6; P < 0.001) and WASO (F (2,12) = 26.2; P < 0.0001). Post-hoc analyses revealed that SOL and WASO were significantly less on NIGHT 2 compared with NIGHTS 1 and 3. No differences in sleep were found between NIGHTS 1 and 3.
Complete subjective sleep data were obtained from eight subjects (Table 1). No significant differences in SSOL, SWASO, or SINSOMNIA were found between treatments (paired t-tests > 0.05). Significant NIGHT effects were found for SWASO (F (2,6) = 4.1; P = 0.04) and SINSOMNIA (F (2,6) = 5.1; P = 0.02); however, post-hoc analyses failed to statistically delineate these main NIGHT effects.
There were no significant correlations between circadian phase-shifts following the conditions and any of the objective or subjective sleep variables on NIGHT 2 or NIGHT 3. There were no treatment effects on sleep with control for circadian phase-shifts by ANCOVA.
Because of technical difficulties, Tre data were obtained from only 11 subjects. Rectal temperature at bedtime during NIGHT 2 was significantly higher (t(11) = 2.34; P = 0.05) following Exercise + Bright Light (37.3 ± 0.5 SE °C) compared with Bright Light Alone (36.9 ± 0.4 SE °C). The decline in Tre from bedtime to the minimum was significantly greater (t(11) = 3.98; P = 0.003) following Exercise + Bright Light (0.88 ± 0.13°C) compared with Bright Light Alone (0.34 ± 0.10°C). Mean rectal temperature during NIGHT 2 did not differ between conditions (Exercise + Bright Light = 36.7 ± 0.5°C; Bright Light Alone = 36.7 ± 0.3)°C. There were no significant correlations among any of these Tre variables and any of the sleep variables. Moreover, ANCOVA, controlling for circadian phase shifts, core temperature at bedtime, and mean temperature during the night showed no significant differences in any of the objective or subjective sleep variables with exercise.
Two women were also tested in the experiment. They also showed no differences in sleep between treatments. For example, mean SOL, WASO, and TST on NIGHT 2 were 8.5 min, 36.5 min, and 194.5 min, respectively, following Exercise + Bright Light and 10.0 min, 29.0 min, and 199.5 min, respectively, after Bright Light Alone.
The primary finding of this study was that the addition of exercise to a 3 h exposure to bright light, ending 30 min before bedtime, did not alter sleep. These data are inconsistent with the teaching that vigorous exercise completed shortly before bedtime delays sleep onset and disrupts sleep (1,2,20). The sleep data on NIGHT 2 are atypical because the sleep period was restricted to 4 h. However, there is little theoretical rationale and no experimental evidence that sleep impairment following exercise would be more profound after more than 4 h of recovery, so it is unlikely that the primary finding of this study would have changed if the subjects had been allowed to sleep for a longer period of time.
The results may have practical applications. For example, following air travel from the East Coast to the West Coast, one might want to exercise vigorously at 11:00 p.m. (2:00 a.m. East Coast time), perhaps to induce a circadian phase delay (34). Many students may also want to exercise after studying until the early morning. Whereas current thinking is that such exercise would disrupt sleep, the results of this study suggest that this may not be so. Because bedtime was delayed by approximately 4 h, it is possible that any sleep-impairing effects of exercise may have been overridden by increased sleep propensity caused by homeostatic and circadian effects. Indeed, on NIGHT 2 both SOL and WASO decreased substantially following both treatments. However, no significant differences in SOL or WASO existed between the treatments, indicating that exercise did not effect this increase in sleep propensity.
Sleep was not assessed polysomnographically, but rather with the Actillume. Numerous research groups have reported very high correlations between actigraphically assessed sleep and polysomnography (PSG) (15,18,21,24,26,30,36). The correlations between PSG and actigraphic assessments of TST have been especially well-documented, but high correlations for SOL, WASO, and sleep efficiency have also been reported.
One limitation of actigraphy is that it cannot distinguish between stages of sleep (30,32). Thus, subtle disruptions in sleep, such as more time spent in lighter sleep stages, may not have been detected in the present study. Whatever differences in sleep that might have occurred were not reflected in the subjective sleep measures.
It is recognized that actigraphy overestimates sleep for individuals who lay motionless while awake (30,32). Thus, it is possible that the Actillume assessment of sleep-wake may have been confounded by lower overall body motility following the exhaustive Exercise + Bright Light condition compared with Bright Light Alone. However, there is no evidence that acute exercise significantly influences body movements during sleep (4,5,6,8,9,11,12,16,38).
Error in actigraphic assessment of sleep is further introduced by individual differences in body movements, both during sleep (e.g., REM-related movements) and during wakefulness (30,32). However, these differences may be less important in a within-subjects study. It has been shown that the sensitivity of actigraphy is sufficient to detect small differences in sleep following experimental manipulations (7,17,31). Nonetheless, the influence of exercise close to bedtime must be assessed with PSG to establish definitively whether there are negative sleep consequences (e.g., more time spent in lighter sleep stages).
Our results are consistent with a recent study (28) which also failed to find any differences in sleep following exercise (60 min at 60% of maximal capacity) compared with sleep following a sedentary control condition completed 30 min before bedtime. Our data suggest that even more prolonged and vigorous exercise (180 min at ∼70% of maximal capacity) near bedtime may have little effect on sleep. The studies are also consistent in showing that Tre was significantly elevated at bedtime following exercise compared with a control condition, but temperature elevation was not associated with sleep. The present study further indicates that neither mean temperature during the sleep period, nor the decline in Tre during the sleep period were associated with sleep. Murphy and Campbell found that sleep onset occurred ∼45-60 min after the maximal rate of decline in Tre at night, depending upon conditions (27). Moreover, they and other researchers have suggested that there is an association between temperature downregulation and sleep, which might be potentiated by thermogenic stimuli such as exercise (10,22,25,27). Our data are not directly comparable with those of Murphy and Campbell because we did not measure the rate of temperature decrease in Tre. However, we can infer that the rate of decrease was greater following Exercise + Bright Light because of the higher bedtime Tre levels and the greater decline in Tre compared with the Bright Light Alone condition. Our data suggest that temperature downregulation following exercise may not be strongly associated with sleep, at least when bedtimes are 30 min after exercise.
The sleep data were apparently also not influenced by circadian rhythm phase-shifts elicited by the treatments. Neither Spearman rank-order correlations between phase-shifts and sleep, nor ANCOVA, controlling for phase-shifts, revealed any significant influence of phase-shifts on sleep. Moreover, evidence indicates that evening bright light does not disrupt sleep when it is administered immediately before (1 min) bedtime (14).
In summary, the findings that an extremely strong exercise stimulus (3 h at 70% of maximal capacity) ending 30 min before bedtime failed to disrupt sleep results suggest that prior assumptions about the sleep-disturbing effects of late-night exercise may not be valid. The ability to sleep soon after vigorous exercise may be unique to physically fit individuals since improvements in fitness can result in quick recovery of sympathetic nervous system arousal following exercise.
1. Sleep Hygiene: Behaviors That Help Promote Better Sleep.
Rochester, MN: American Sleep Disorders Association, 1997, p. 2.
2. Ancoli-Israel, S. Sleep Disorders: All I Want is a Good Night's Sleep.
Chicago: Mosby-Year Book, 1996.
3. Ashton, H. Toxicity and adverse consequences of benzodiazepine use. Psychiatr. Ann.
25:158-166, 1995, p. 46.
4. Baekeland, F. Exercise deprivation: sleep and psychological reactions. Arch. Gen. Psychiatr.
5. Bevier, W. C., D. L. Bliwise, N. G. Bliwise, D. E. Bunnell, and S. M. Horvath. Sleep patterns of older adults and the effects of exercise. J. Clin. Exper. Gerontol.
6. Bonnet, M. H. Sleep, performance, and mood after the energy-expenditure equivalent of 40 hours of sleep deprivation. Psychophysiology
7. Brooks, J. O., L. Friedman, D. L. Bliwise, and J. A. Yesavage. Use of the wrist actigraph to study insomnia
in older adults. Sleep
8. Browman, C. P. Sleep following sustained exercise. Psychophysiology
9. Browman, C. P. and D. I. Tepas. The effects of pre-sleep activity
on all-night sleep. Psychophysiology
10. Bunnell, D. E., J. A. Agnew, S. M. Horvath, L. Jopson, and M. Wills. Passive body heating and sleep: influence of proximity to sleep. Sleep
11. Bunnell, D. E., W. C. Bevier, and S. M. Horvath. Effects of exhaustive exercise on the sleep of men and women. Psychophysiology
12. Bunnell, D. E., W. C. Bevier, and S. M. Horvath. Sleep interruption and exercise. Sleep
13. Buysse, D. J., C. F. I. Reynolds, T. H. Monk, S. R. Berman, and D. J. Kupfer. The Pittsburgh Sleep Quality Index: a new instrument for psychiatric practice and research. Psychiatry Res.
14. Campbell, S. S., D. J. Dijk, Z. Boulos, C. I. Eastman, A. J. Lewy, and M. Terman. Light
treatment for sleep disorders: consensus report III. alerting and activating effects. J. Biol. Rhythms
15. Cole, R. J., D. F. Kripke, W. Gruen, D. J. Mullaney, and J. C. Gillin. Automatic sleep/wake identification from wrist activity
16. Desjardins, J., T. Healey, and R. J. Broughton. Early evening exercise and sleep (Abstract). Sleep Res.
17. Drennan, M. D., D. F. Kripke, H. A. Klemfuss, and J. D. Moore. Potassium affects actigraph-identified sleep. Sleep
18. Dunham, D. W., R. F. Hoffmann, and R. J. Broughton. Wrist actigraphy
and sleep/wake estimation revisited (Abstract). Sleep Res.
19. Hasan, J., H. Urponen, I. Vuori, and M. Partinen. Exercise habits and sleep in a middle-aged Finnish population. Acta Physiol. Scand.
20. Hauri, P. J. Sleep Disorders.
Kalamazoo, MI: Upjohn Company, 1992, pp. 20-21.
21. Hauri, P. J., and J. Wisbey. Wrist actigraphy
22. Horne, J. A., and V. J. Moore. Sleep EEG effects of exercise with and without additional body cooling. Electroencephalogr. Clin. Neurophysiol.
23. Kripke, D. F., M. R. Klauber, D. L. Wingard, R. L. Fell, J. D. Assmus, and L. Garfinkel. Mortality hazard associated with prescription hypnotics. Biol. Psychiatr.
24. Kripke, D. F., D. J. Mullaney, S. Messin, and V. G. Wyborney. Wrist actigraphic measures of sleep and rhythms. Electroencephalogr. Clin. Neurophysiol.
25. McGinty, D., and R. Szymusiak. Keeping cool: a hypothesis about the mechanisms and functions of slow wave sleep. Trends Neurosci.
26. Mullaney, D. J., D. F. Kripke, and S. Messin. Wrist-actigraphic estimation of sleep time. Sleep
27. Murphy, P. J., and S. S. Campbell. Night-time drop in body temperature: a physiological trigger for sleep onset? Sleep
28. O'Connor, P. J., M. J. Breus, and S. D. Youngstedt. Exercise-induced increase in core temperature does not disrupt a behavioral measure of sleep. Physiol. Behav.
29. Sadeh, A., J. Alster, D. Urbach, and P. Lavie. Actigraphically based automatic bedtime sleep-wake scoring: validity and clinical applications. J. Ambul. Monit.
30. Sadeh, A., P. J. Hauri, D. F. Kripke, and P. Lavie. An American Sleep Disorders Association review: the role of actigraphy
in the evaluation of sleep disorders. Sleep
31. Schmidt-Nowara, W. W., A. A. Beck, and C. A. Jessop. Actigraphic assessment of a treatment trial of sleep restriction in chronic insomnia
. Sleep Res.
32. Standards of Practice Committee, A. An American Sleep Disorders Association Report: practice parameters for the use of actigraphy
in the clinical assessment of sleep disorders. Sleep
33. Urponen, H., I. Vuori, J. Hasan, and M. Partinen. Self-evaluation of factors promoting and disturbing sleep: an epidemiological survey in Finland. Soc. Sci. Med.
34. Van Reeth, O., J. Sturis, M. M. Byrne, et al. Nocturnal exercise phase delays circadian rhythms of melatonin and thyrotropin secretion in normal men. Am. J. Physiol.
35. Vuori, I., H. Urponen, J. Hasan, and M. Partinen. Epidemiology of exercise effects on sleep. Acta Physiol. Scand.
36. Webster, J. B., D. F. Kripke, S. Messin, D. J. Mullaney, and G. Wyborney. An activity
-based sleep monitor system for ambulatory use. Sleep
37. Youngstedt, S. D., D. F. Kripke, J. A. Elliott, and J. D. Assmus. Phase shifting effects of exercise combined with bright light
(Abstract). Sleep Res.
38. Youngstedt, S. D., P. J. O'Connor, R. K. Dishman, J. B. Crabbe, and K. I. Shiver. Influence of exercise on caffeine-induced insomnia
(Abstract). Sleep Res.