Daily exercise facilitates phase delays of circadian melatonin rhythm in very dim light

Laura K. Barger, Kenneth P. Wright Jr., Rod J. Hughes, Charles A. Czeisler


Shift workers and transmeridian travelers are exposed to abnormal work-rest cycles, inducing a change in the phase relationship between the sleep-wake cycle and the endogenous circadian timing system. Misalignment of circadian phase is associated with sleep disruption and deterioration of alertness and cognitive performance. Exercise has been investigated as a behavioral countermeasure to facilitate circadian adaptation. In contrast to previous studies where results might have been confounded by ambient light exposure, this investigation was conducted under strictly controlled very dim light (standing ∼0.65 lux; angle of gaze) conditions to minimize the phase-resetting effects of light. Eighteen young, fit males completed a 15-day randomized clinical trial in which circadian phase was measured in a constant routine before and after exposure to a week of nightly bouts of exercise or a nonexercise control condition after a 9-h delay in the sleep-wake schedule. Plasma samples collected every 30–60 min were analyzed for melatonin to determine circadian phase. Subjects who completed three 45-min bouts of cycle ergometry each night showed a significantly greater shift in the dim light melatonin onset (DLMO25%), dim light melatonin offset, and midpoint of the melatonin profile compared with nonexercising controls (Student t-test; P < 0.05). The magnitude of phase delay induced by the exercise intervention was significantly dependent on the relative timing of the exercise after the preintervention DLMO25% (r = −0.73, P < 0.05) such that the closer to the DLMO25%, the greater the phase shift. These data suggest that exercise may help to facilitate circadian adaptation to schedules requiring a delay in the sleep-wake cycle.

  • circadian
  • jet lag
  • shift work

scheduled physical activity such as wheel running in rodents and exercise in humans has been reported to influence the circadian timing system. Wheel running (12, 31) and forced treadmill running (21, 24) have been reported to entrain circadian rhythms in hamsters (31) and mice (12, 21) and when the period of the free-running rhythm was close to the period of the treadmill schedule in rats (24). Several studies have assessed the effectiveness of exercise as a circadian phase-resetting agent in humans. Van Reeth et al. (34) reported that a single exercise bout, centered from 3 h before to 2 h after the body temperature minimum, phase delayed the human circadian rhythms of temperature and thyrotropin. In that study, exercise that consisted of 3 h of alternating arm and leg ergometry at 40 and 60% of maximal O2 consumption (V̇o2 max) was performed under constant lighting conditions of <300 lux. Using the same protocol, Baehr and colleagues (1) reported similar phase delays between young and older adults. Another study (7) reported that higher-intensity exercise of shorter duration, 1 h of stair climbing at 75% V̇o2 max, centered at 0100 (in ∼70–80 lux), also elicited significant phase delays in the circadian rhythm of thyrotropin. Recently, this protocol was expanded to include scheduled morning, afternoon, evening, and night exercise sessions in ∼40 lux (8). Significant phase delays were reported for noctural exercise, and significant phase advances were reported for evening exercise. Only one other study to date has reported that exercise can advance the human circadian pacemaker. In a shortened T-cycle protocol of twelve 23-h and 40-min cycles, subjects who performed two bouts of cycling and rowing at a heart rate of 140 beats/min during the morning and afternoon showed a 1.6-h advance in the peak of the plasma melatonin rhythm, whereas subjects in the nonexercise control condition averaged a 0.80-h phase delay of the melatonin peak (25).

Exercise has also been reported to facilitate adaptation to nightshift work in a field study that exposed subjects to a 9-h phase delay and daytime sleep (11). Investigations simulated lighting conditions of a typical night shift in ordinary room light of <500 lux. Subjects assigned to the exercise group cycled on a stationary ergometer for 15 min, at 50–60% maximum heart rate, every hour during the 8-h shift on the first three nightshifts. With the use of the timing of demasked body temperature minimum to determine phase shifts, results indicated 63% of the exercise group shifted >6 h, whereas only 38% of the control group showed shifts that large. When differences in morningness/eveningness scores between groups were accounted for, the differences in phase shifts were statistically significant.

However, in the above studies, it is unclear to what extent exercise was responsible for phase shifting the circadian pacemaker because subjects were exposed to light levels that were known to phase shift the human circadian pacemaker (5, 6, 41). Two studies have examined the interaction of exercise and bright light on circadian rhythms in humans. Baehr et al. (2) investigated the interaction of bright light (>5,000 lux) or room light (<500 lux) with or without six bouts of 15 min cycle ergometer exercise at 50–60% of maximum heart rate in a simulated night work protocol with a 9-h phase delay in the light-dark cycle. Analysis of demasked body temperature rhythm showed no difference in the number of hours shifted between the exercise and control groups. Exercise neither facilitated nor inhibited the shifting effects of bright light. The authors hypothesized that the low intensity and short duration of the exercise in their study were insufficient to elicit phase shifts of the endogenous circadian pacemaker. Most recently, Youngstedt et al. (40) used a within-subjects counterbalanced design to compare the phase-shifting effects of 3 h of bright light (3,000 lux) alone and in combination with 3 h of cycle ergometry at 65–75% of heart rate reserve. Illumination was maintained at 50 lux during wake times. A significant phase delay in the rectal temperature minimum was reported after the bright light alone condition, although no difference was reported for the acrophase of urinary 6-sulphatoxymelatonin (6-SMT). The combination of bright light and exercise elicited a significant phase delay in 6-SMT, but the delay was not significantly different from that produced by bright light alone. The authors concluded that their study did not have adequate statistical power to detect a phase shift difference of <2–2.5 h between conditions.

As noted, the aforementioned studies are limited since they did not adequately control for the phase-resetting effects of light on the circadian system. Therefore, the current study tested the effectiveness of moderate exercise to phase delay the human circadian pacemaker under very dim light conditions.



Eighteen healthy, physically fit, nonsmoking young males [23.0 ± 3.6 (SD) yr] completed a 15-day inpatient protocol. Participants each gave informed consent in writing. The Brigham and Women's Hospital/Partners Health Care System Human Research Committee approved the procedures for the protocol. The investigation was conducted according to the principles expressed in the Declaration of Helsinki and the Belmont Report. Each participant underwent an extensive screening process and was free from any acute, chronic, or debilitating condition. Health evaluation was based on clinical history, electrocardiogram (ECG), clinical biochemical screening of blood and urine, psychological questionnaires, and psychological and physical examination. Each subject was administered the Horne-Ostberg questionnaire to establish morningness-eveningness preference (16), and anyone who scored as an extreme morning or extreme evening type was excluded. Toxicology screens for drug use verified that participants were drug free, including caffeine and alcohol, upon admission to the laboratory. Subjects were ineligible for the study if they had performed shift work in the last 18 mo or traveled across two or more time zones during 3 mo before the study.

Physical fitness was verified by a symptom-limited maximal exercise test given during the screening process. This test was administered on a treadmill using a Bruce protocol that began at 1.7 miles/h (mph) and 10% grade, and speed was increased by ∼0.8 mph and grade by 2% every 3 min. Heart rate, blood pressure, and ECG were monitored throughout the treadmill test. The protocol continued until the subject fatigued and indicated he could not continue. None of the following potential endpoints were observed: abnormal ECG, decreased systolic blood pressure; angina or dyspnea, and no adverse events. Subjects that were able to complete 12 min of the treadmill protocol were eligible for this study.

Ambulatory physiological monitoring.

Subjects were required to maintain a regular, self-selected sleep-wake schedule of ∼8 h sleep/night for 3 wk before admission to this study. Sleep-wake diaries and call-in times to a date- and time-stamped answering machine just before going to bed and immediately upon awakening were used to monitor compliance with this protocol. In addition, wrist activity and ambient light levels were monitored for 1 wk immediately before admission to the laboratory with a solid-state, portable data collection device (Actiwatch-L; Minimitter, Bend, OR). If sleep or wake times differed by >30 min on more than two occasions, the subject was disqualified from participation.

Inpatient protocol.

The laboratory portion of the protocol (Fig. 1A) began with three acclimation days and nights with 16 h of scheduled wakefulness and 8 h of scheduled sleep at the subject's habitual time. This was followed by a 49-h constant routine (CR) protocol (described below) to assess circadian phase. The length of the CR resulted in a 9-h delay of the subject's habitual sleep episode and thus a 9-h delay of their sleep-wake cycle. This delayed schedule was maintained for seven experimental days.

Fig. 1.

A: 15-day experimental laboratory protocol. Sleep episodes (8 h) are shown as solid bars. During the first baseline day, the subject was exposed to 150 lux (gray bars). For the remainder of the study, subjects were in the <5 lux condition (open bars). After the 3 baseline days, there was a 49-h constant routine (CR; hatched bar). The subject's sleep-wake schedule was then delayed by 9 h. During the 7 experimental days of the protocol, the 45-min exercise/control bouts occurred nightly (▾) and were centered 10.875 h after the new scheduled wake time. A second CR (40 h) was scheduled before the last sleep episode. B: 3 45-min bouts of exercise or the control condition (▾) separated by 1-h rest periods were performed daily throughout the experimental portion of the protocol. The first bout began 8.75 h after waking and ended 9.50 h after wake time, the second from 10.50 until 11.25 h after wake time, and the third from 12.25 until 13.00 h after wake time. The final session ended 3 h before the beginning of the sleep period.

We chose to perform a between-subjects comparison rather than a crossover design to avoid potential aftereffects to a phase-shifting protocol that have been reported to occur in animals (20, 28). Thus each participant was randomly assigned to either a control or exercise group (n = 9 in each group). From day 7 to 13, subjects in the exercise group underwent three 45-min bouts of cycle ergometry, whereas subjects in the control group sat on the cycle without peddling. A second CR of 40 h in duration was used to assess circadian phase after the experimental treatment days.

Inpatient environment and conditions.

Upon admission to the study, subjects were maintained in an environment free of time cues, including clocks, radios, television, visitors, and sunlight but maintained contact with staff members using techniques described elsewhere (35). Environmental temperature was maintained at 24 ± 2°C.

During scheduled wakefulness, subjects were free to move about the suite as desired, except to lie down or nap. Wakefulness and activity were monitored via closed-circuit cameras and frequent interaction with research staff. Exercise outside of the allocated times was prohibited. During scheduled sleep, subjects were instructed not to get out of bed, even if they awakened before the end of the scheduled sleep episode. If requested by the subject, a technician brought the subject a urinal or bedpan during the scheduled sleep episode.

The CR consisted of a regimen of enforced semirecumbent wakefulness in constant dim light. Nutritional intake was divided into hourly aliquots, and activity was restricted to prevent changes in body posture and activity level. A technician was present at all times during the CRs to ensure wakefulness.

Light exposure.

Ceiling-mounted fluorescent lamps (T8 and T80 lamps; Philips Lighting, Eindhoven, NE) with a 4100K color temperature produced a spectrum of white light. Light exposure during scheduled wakefulness on the three baseline study days was ordinary indoor room light with intensity of ∼150 lux (measured horizontally at ∼183 cm). During the CRs and experimental portions of the study (days 7–13), light exposure was very dim. Maximum ambient light exposure during these portions of the study was <5 lux, measured at ∼183 cm from the floor in the direction of the ceiling fixtures. The average light intensity in the standing angle of gaze measured horizontally at eye level during scheduled wakefulness of the experimental portion of the protocol was 0.65 ± 0.02 lux. Participants were scheduled to sleep in darkness. Light intensity was measured with an IL-1400 photometer (International Light, Newburyport, MA). The dim light intensities during scheduled wakefulness permitted the assessment of melatonin levels without concern over the acute melatonin suppressant effects of light (42).

Hormonal data.

Blood samples were collected every 30 min throughout the CRs and every 60 min during other portions of the protocol through an indwelling 18-gauge intravenous catheter located in a forearm vein. The catheter was connected to a triple-stopcock manifold (Cobe Laboratories, Lakewood, CO) via an intravenous loop with a 12-foot small-lumen extension cable (Sorex Pharmaceuticals, Salt Lake City, UT) through which blood sampling continued in the next room during sleep. Between samples, a solution of 0.45% saline with 10,000 IU/l heparin was infused at a rate of 20 ml/h to maintain patency. Blood samples were transferred to small (3-ml) vacutainer tubes and immediately centrifuged at 4°C; the resulting plasma was pipetted into polystyrene tubes and frozen at −20°C until analysis. Plasma melatonin was assayed using an 125I RIA technique (Diagnostic, Osceola, WI). The sensitivity of the assay was 2.5 pg/ml. The average interassay and intra-assay coefficients of variation were 7.38 and 6.88%, respectively.

Exercise intervention.

In subjects randomized to the exercise group, three 45-min bouts of exercise separated by 1-h rest periods were performed daily from day 7 to day 13 of the protocol. The first exercise bout began 8.75 and ended 9.50 h after waking. There was then a 1-h rest period (during which the subject was free to move about the suite but not allowed to lie down), followed by exercise bout 2 from 10.50 until 11.25 h after the scheduled wake time. Another 1-h rest period then occurred, followed by exercise bout 3, from 1225 until 1300 after wake time. The duration of the exercise intervention, including rest periods, was 4.25 h. Exercise was centered in the latter part of the waking day and, because of the delay in the sleep-wake cycle, occurred during the subject's biological night (Fig. 1B).

During each exercise bout, subjects were required to maintain an intensity of 65–75% of their maximal heart rate (MHR) that was determined during the symptom-limited maximal exercise test. To achieve this rate, subjects pedaled at 65–70 revolutions/min (rpm) on a bicycle ergometer (Cybex model 700R; Cybex International, Medway, MA). This is considered a moderate to hard level of exercise (29). Analysis of heart rate during the exercise sessions indicated that the average percentage of maximal heart rate maintained throughout all bouts of exercise was 69 ± 6% (mean ± SD). A trained technician was present to oversee each exercise bout, to measure blood pressure, to record heart rate and ergometer rpm every minute, to record perceived rate of exertion, and to ensure that the target exercise intensity was achieved. Subjects performed leg-stretching exercises before and after each exercise intervention. Subjects randomized to the control group underwent the same protocol, including pre- and postexercise stretching procedures, at the same scheduled time after waking and for the same durations. However, control subjects only sat on the ergometer to control for posture but did not pedal. A technician was present to supervise all control sessions.

Statistical analysis.

The time at which the rising portion of the plasma melatonin curve crossed a level that was 25% of the three-harmonic fitted peak to trough amplitude of the curve during the first CR was defined as the dim light melatonin onset (DLMO25%; see Refs. 37 and 38). The dim light melatonin offset (DLMOff25%) was defined as the time at which the falling portion of the interpolated plasma melatonin curve crossed a level that was 25% of the three-harmonic fitted peak-to-trough amplitude of the curve. The melatonin midpoint between the DLMO25% and the DLMOff25% was defined as the midpoint of the melatonin curve.

Melatonin phase was analyzed for each day of the protocol during and between the CRs. Student's t-tests were employed to compare DLMO25% and phase angle on CR1 between exercise and control groups and to compare the difference in melatonin phase between the pre- and postexperimental CRs in the exercise and control groups.

Daily melatonin phase estimates provided analysis of the day-to-day progression of phase shifts in all subjects. Repeated-measures ANOVA with modified Bonferroni correction factors for multiple comparisons were used to test for differences across days (within-subject factor) and between conditions. Huynh-Feldt degrees of freedom correction factors were used to correct for the violation of sphericity assumption. Pearson production correlation (Pearson r) was used to determine the association between the timing of exercise and DLMO25%. All analyses were performed using Statistica (StatSoft, Tulsa, OK).


Initial circadian phase, as assessed by the DLMO25% during CR1, was similar for those assigned to the control [21.34 ± 1.46 (SD) h] and to the exercise [22.15 ± 1.58 (SD) h] conditions [t = 1.13; degrees of freedom (df) = 16, P = 0.28]. Additionally, there was no statistically significant difference between the exercise (47.7 + 7.1, mean + SD) and control (50.3 + 7.7, mean + SD) group in morningness-eveningness scores as measured by the Horne-Ostberg questionnaire (t = 0.75, df = 16, P = 0.47; see Ref. 16).

Data from the pre- and postexperimental intervention CRs were assessed to determine whether repetitive nightly bouts of exercise elicited significant phase delay shifts of the endogenous circadian melatonin rhythm. Analysis of the DLMO25% revealed a significantly greater phase delay in the exercise subjects [3.17 + 0.49 (SE) h] compared with the control subjects [1.67 + 0.45 (SE) h; Fig. 2A; t = 2.25, df = 16, P = 0.039]. Likewise, the midpoint (Fig. 2B; t = 2.44, df = 16, P = 0.027) and the DLMOff25% (Fig. 2C; t = 2.57, df = 16, P = 0.021) of the plasma melatonin rhythm were significantly phase delayed in subjects exposed to the exercise vs. the control condition [3.34 + 0.52 vs. 1.59 + 0.50 (SE) h and 3.51 + 0.55 vs. 1.51 + 0.55 (SE) h, respectively].

Fig. 2.

Dim light melatonin onset (DLMO25%; A), the melatonin midpoint (B), and dim light melatonin offset (DLMOff25%; C) were compared between preexperimental (CR1) and postexperimental (CR2) intervention CRs. Symbols represent individual subjects; the short lines represent the mean with SE. The exercise group showed a significantly greater phase delay compared with the control group. *P ≤ 0.05.

Repeated-measures ANOVA revealed that exercise significantly delayed the melatonin phase compared with controls. There were significant interaction effects for days since CR1 by condition for midpoint (F = 4.03, df = 2.3, 36.3, epsilon = 0.28; P = 0.03) and DLMOff25% (F = 3.24, df = 4.3,68.3, epsilon = 0.53; P = 0.03). The interaction effect for days since CR1 for the DLMO25% by condition (F = 3.18, df = 1.7, 28.4, epsilon = 0.22; P = 0.06) showed a trend toward significance. Planned comparisons with Bonferroni corrections revealed that, on day 9 after CR1, the DLMO25% was significantly different between the exercise and control group (Fig. 3A). Furthermore, on days 5–9, the midpoint (Fig. 3B) and DLMOff25% (Fig. 3C) were significantly delayed by exercise. The melatonin phase change observed in the control group was consistent with the period-dependent phase drift expected in dim light (0.2-h delay/day, based on the average intrinsic circadian period of 24.2 h; see Ref. 38). None of the subjects rapidly adapted to the shift in the sleep-wake cycle (30).

Fig. 3.

The cumulative melatonin phase shift produced in the exercise group (•) across days since CR1 was significantly greater than that seen in the control group (○). Planned comparisons revealed significant differences between groups. The straight line for DLMO25% (A) represents the expected change in phase (0.2-h delay/day) if subjects drifted at the average intrinsic circadian period (24.2 h; see Refs. 9 and 38). *P ≤ 0.044.

Because this protocol scheduled the midpoint of the experimental intervention 19.88 h after habitual wake time (10.75 h after the delayed-schedule wake time) and not according to endogenous circadian time, there were individual differences in the endogenous circadian phase that exercise occurred. Exercise was centered 4.2–6.7 h after DLMO25% for exercise subjects, and sitting on the bike was centered 4.7–7.4 h after DLMO25% for control subjects. We observed that the amount of phase delay induced by the exercise intervention was significantly dependent on the relative timing of the exercise with respect to the preintervention DLMO25% (r = −0.73, P < 0.05). Figure 4 shows the relationship between the relative timing of the intervention [control (A) and exercise (B)] with respect to the preintervention DLMO25% and the magnitude of the phase shift produced by the intervention. Exercise centered closer to the preintervention DLMO25% elicited larger phase delays. The timing of the control session, with the subject sitting on the bike but not pedaling, and the amount of drift in circadian phase was not significantly related (r = −0.28).

Fig. 4.

The amount of phase shift achieved was significantly dependent on the timing of exercise (B, •; r = −0.73, P = 0.025). This relationship was not significant for the control intervention (A, ○; r = −0.28, P = 0.465).


Exercise facilitated the shift of the timing of the circadian clock to a 9-h delay of the sleep-wake cycle. The melatonin rhythm of the exercise group showed significant phase delays in the DLMO25%, the midpoint, and the DLMOff25% relative to the nonexercising control condition. Although none of the subjects was able to completely adapt to the 9-h shift in the light-dark cycle, our results represent the first demonstration of exercise facilitating a phase shift while strictly controlling for the effects of light by keeping subjects in near dark conditions. It was anticipated that, over the nine intervening 24-h cycles between the pre- and post-CR assessments in dim light, the endogenous circadian pacemaker of the nine subjects studied under the control condition would drift with an average intrinsic period of ∼24.2 h, which would result in an observed phase delay shift of the endogenous circadian melatonin rhythm of ∼1.8 h (9). Indeed, the mean shift of the control group closely matched our expectations (38).

The exercise group showed a significant relationship between DLMO25% on CR1 and the phase delay in DLMO25% between CR1 and CR2, but this relationship did not hold true in the control group. The timing of the exercise in relationship to the circadian phase angle was significantly correlated with the size of the delay shift achieved. This finding is consistent with previous research indicating that the sensitivity of the human circadian timing system to exercise is not constant across time of day (7, 8, 34). That is, the direction and magnitude of a shift caused by exercise is dependent on the internal biological time that the stimulus is presented.

In the present study, the greatest shifts in melatonin onset occurred when exercise was centered ∼4 h after DLMO25%. There was a significantly linear decrease (r = −0.73; P < 0.05) in the phase shift achieved as the stimulus was applied at a later circadian phase. Although the range of our stimulus application was very limited (4–7 h after DLMO25%), our data appear to correspond temporally to the partial exercise-phase response curves previously developed (7, 34) and to the full exercise curve more recently constructed by Buxton et al. (8), even though those studies were conducted at a much higher light intensity (8).

There was no significant relationship between timing of the control stimulus and the amount of phase delay achieved in control subjects (r = −0.28). Thus the association between the timing of DLMO25% and the phase shift produced was not merely the result of the change in the sleep-wake cycle or the dim light-dark cycle, but specifically related to exercise. Because the timing of the experimental intervention was linked to wake time, those subjects with later melatonin phases received the exercise stimulus at an earlier relative time after DLMO25%. It is possible that different melatonin phases, and perhaps associated intrinsic periods (10), could be responsible in part for the variation in phase delays produced. Because the exercise and control groups were not significantly different with regard to the timing of their initial DLMO25% or their morningness-eveningness scores, this explanation seems unlikely.

Whereas our results are qualitatively similar to the timing in the previously published one-pulse phase-response curves to exercise (7, 34), the magnitude of the response is not consistent. Because our subjects exercised in the middle of the biological night, one would expect that the seven consecutive days of exercise in this region of the phase-response curve would elicit a greater response than a single exercise episode. Both Van Reeth et al. (34) and Buxton et al. (7, 8) showed phase delays of up to 3 h after a single bout of exercise of similar duration. The present study demonstrated delays of 0.95–5.45 h, with an average of 3.17 h in DLMO25% after a week of exercise during the biological night.

The phase assessment protocol used in previous studies (7, 34) examined circadian phase on the same day that the exercise stimulus was presented. This procedure makes it difficult to evaluate whether the reported phase shifts were sustained or the result of an acute effect of exercise. This point is illustrated by recent work from Buxton and colleagues (8) that included a measurement of melatonin phase on the day of and the day after the exercise stimulus. When compared with the melatonin phase on the day before the exercise stimulus, melatonin phase shifts were significantly different the day of the stimulus but not always on the day after the stimulus. Our day-by-day analysis revealed that the differences in melatonin phase between the exercise and control groups were not statistically different in the initial days of the experimental portion of the protocol. We believe the assessment of circadian phase on multiple days during and under constant conditions after the exercise stimulus is more appropriate to demonstrate reliable and sustained phase shifts in the melatonin rhythm.

Lighting conditions in the previous studies may also account for the larger magnitude of shifts after a single pulse of exercise. Previous exercise studies have been conducted under a variety of lighting conditions. Van Cauter et al. (33) and Van Reeth et al. (34) reported that one bout of exercise elicited 1- to 2-h phase delays in plasma melatonin and thyrotropin in a background of <300 lux. Eastman and colleagues (11) exposed subjects to <500 lux during exercise in a study where subjects cycled 15 min/h (50–60% maximum heart rate) for 8 h on three consecutive nights and reported a greater number of large phase shifts in demasked body temperature (>6 h) than nonexercising control subjects. Buxton et al. (7) reported phase delays of up to 2 h after a single bout of exercise in lighting conditions of 70 and 80 lux and more recently at an average of 42 lux (8). Because it has been shown that dim light of <12 lux can significantly reset the endogenous circadian rhythm of plasma melatonin (37) and that one-half of the maximal phase-delaying response to ∼10,000 lux can be achieved with 100–200 lux (41), it is likely that the brighter lighting used in previous exercise studies influenced the phase shifts reported.

However, recent studies failed to show a significant interaction between bright light and exercise. Using demasked body temperature as a marker of the circadian system, Baehr et al. (2) concluded that exercise neither facilitated nor inhibited phase shifts produced by bright light (∼5,000 lux). The results of their study, however, may be compromised by the intensity of the “dim light” (<500 lux), the low level of exercise (15-min bouts of cycling at 50–60% of maximum heart rate, hourly for 6 h), or the demasking method used to assess the phase shift (19). Youngstedt et al. (40) also reported no difference between phase shifts resulting from bright light alone (3 h of 3,000 lux with background lighting of 50 lux) and bright light combined with 3 h of cycling at 65–75% of heart rate reserve [approximately equivalent to 72–81% MHR (23) when measuring circadian phase using body temperature minimum and 6-SMT acrophase as circadian markers]. The later negative finding could be because of less sensitive phase markers of the circadian system or, as the authors assert, inadequate statistical power. Despite the results of these two trials, our study suggests that a possible synergistic effect of light with exercise might exist because in the near-dark conditions of this study, the magnitude of phase delays elicited by exercise is relatively small compared with previous studies conducted in brighter light. Future phase-response curves should be conducted in very dim light to clarify the phase-resetting capacity of exercise.

It is well established that the intensity, duration, and biological time of exposure to environmental time cues are important in determining the phase-resetting response of the circadian timing system. The partial exercise phase-response curves were created with 1 h of intense exercise on the stair master (25 and 75% V̇o2 max; equivalent to ∼48 and 84% MHR; see Ref. 23) and 150 min of exercise over 3 h on a cycle/arm ergometer at a lesser intensity (alternating 40 and 60% V̇o2 max; equivalent to ∼59 and 73% MHR; see Refs. 7 and 23) at different circadian phases. Another study using low-intensity exercise (15 min cycling/h for 8 h at 50–60% MHR) produced less reliable phase delays.(11). Our study used three 45-min bouts of moderate-intensity exercise on a cycle ergometer (MHR 69 ± 6%, mean ± SD) and thus was similar in intensity and duration to the previous cycle/arm ergometer study (7). The optimal intensity and duration of exercise to facilitate phase shifting has yet to be determined.

It is theoretically possible to achieve the observed phase shift through transient changes in intrinsic period. However, the evidence to date does not show an influence of exercise on intrinisic period in humans, nor has any environmental time cue been shown to alter intrinsic period in humans (3, 32). However, activity has been reported to be associated with shorter intrinsic periods in rats (39) and mice (13) and longer intrinsic periods in hamsters (26) in studies performed under constant conditions. Furthermore, the timing, quantity, and intensity of activity have been reported to influence the intrinsic period of rodents (14, 27). Additional human studies varying the timing, quantity, and intensity of exercise under forced desynchrony or constant conditions will have to be conducted to examine the influence of exercise on intrinsic period in humans.

Our study attempted to control for the photic pathway in phase shifting by maintaining the subjects in both the control and exercise groups in very dim light. If the exposure to dim light in our study made the circadian system more sensitive to light exposure, the effect would have to be such that an average of ∼0.65 lux (angle of gaze) would have sufficient phase-shifting effects only in the exercise group. This may have been possible if the exercise elicited increased pupillary dilation, as that has been reported to yield greater phase shifts in response to light (15). However, there is no evidence that exercise of the intensity we used in this study influences pupil size (17, 36); thus, this is a highly unlikely factor in the possible explanations of differences between groups.

Our results are the most convincing to date to suggest a nonphotic pathway that facilitates phase delays of the human endogenous circadian pacemaker. Animal models implicate the intergeniculate leaflet is anatomically important in the transfer of nonphotic information to the suprachiasmatic nucleus (SCN) via neuropeptide Y (4, 18). Other research suggests that the serotonergic system may also be involved in nonphotic input to the SCN (22). Exercise produces a multitude of physiological responses in humans, including hormonal changes and increased body temperature. Future research will be needed to fully understand the mechanism that constitutes the nonphotic cue and the transfer of that information to promote circadian phase shifting.

In summary, these data suggest that exercise can significantly phase delay the human circadian pacemaker and may help to facilitate circadian adaptation to schedules requiring a delay in the sleep-wake schedule. The optimal duration and intensity of exercise for maximal phase shifting of the human circadian timing system have not been described. Further studies using human subjects are needed to determine the factors (duration, intensity, modality, and timing) of human exercise that best facilitate phase shifting or entrainment. Complete phase-response curves to exercise, both in the presence and in the absence of light, should be constructed. Our finding that exercise shifts are still possible in near-dark conditions may have implications for entraining or phase shifting blind individuals who do not have light input in their SCN. The results may also have important implications for the use of exercise to facilitate adaptation to shift work schedules and non-24-h schedules such as those required for long-term space exploration.


This work was supported by grants from the United States Air Force Office of Scientific Research (F49620-00-1-0334), the National Aeronautics and Space Administration Cooperative Agreement NCC9–58 with the National Space Biomedical Research Institute, and the Training Program in Sleep, Circadian, and Respiratory Neurobiology [National Institutes of Health (NIH) T 32 HL-07901]. Brigham and Women's Hospital General Clinical Research Center was supported by NIH grant M01-RR-02635.


We acknowledge the contributions of David Rimmer in the development of this study and work with the initial subjects of the protocol.


  • The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.


View Abstract