INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

IT HAS BEEN LONG RECOGNIZED that the alternation of light and dark plays a dominant role in the synchronization of physiological processes controlled by the central circadian pacemaker. In the human, the majority of studies conducted over the last two decades of the synchronization of circadian rhythms, such as the sleep-wake cycle and neuroendocrine secretory profiles, has focused on the role of light exposure (4). In contrast, little attention has been given to the possible role of dark exposure and the typically associated state of rest or sleep, referred to as dark/sleep in the remainder of this article. The vast majority of studies examining strategies to facilitate the realignment of circadian rhythms using manipulations of the photic environment in nonnatural modern situations, such as night work or transmeridian air travel, has also focused on the timing, intensity, and duration of light exposure. Whether appropriately timed exposure to darkness and/or sleep may serve as a temporal signal for the human circadian clock remains largely unknown. In addition to its theoretical importance, this question has practical implications for shift workers and for those experiencing jet lag, because, in these subject populations, the misalignment between endogenous circadian rhythmicity and environmental or social periodicities often results in a need to be active at a time of minimum physiological vigilance and resting at a time of minimal sleep propensity, with deleterious consequences for health, safety, and productivity.

Extensive data indicate the existence of a nonphotic input to the mammalian circadian clock (15, 16). In particular, a variety of nonphotic stimuli that involve changes in the rest-activity state may induce phase shifts of circadian rhythms. In nocturnal rodents, a period of darkness presented over a background of constant light induces increased activity and phase shifts of locomotor rhythms (10). Such phase shifts are most robustly elicited when timed to occur during a normal period of sleep and/or low activity levels. The recent characterization of a phase response curve to dark pulses in a diurnal rodent (13) suggests that dark pulses can similarly phase shift circadian rhythms in a diurnal species, raising the possibility that dark pulses could phase shift rhythms in the human. Existing evidence supporting a role for nonphotic influences on the human circadian system includes studies that found phase shifts after exposure to nocturnal exercise in normal adult subjects (2, 20), as well as observations of entrainment of circadian rhythmicity to 24-h and non-24-h cycles in blind subjects receiving no light/dark cues (12). Furthermore, several studies have indicated that the timing of the sleep/dark period can modulate the response of the circadian system to a given light stimulus (7). Finally, two recent reports examining phase shifts of hormonal markers of circadian phase in response to an 8-h advance (17) or 8-h delay (11) of the sleep period suggest that daytime sleep in darkness may be able to phase shift the human circadian clock.

The purpose of this study was to systematically determine the effects of daytime sleep in darkness on the phase of the human circadian system, using two established hormonal markers of human circadian phase (1, 2, 11, 20).

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects. Twenty-five normal young men aged 20-30 yr (mean age ± SD = 23 ± 3 yr) with normal body weight (mean body mass index ± SD = 24.0 ± 2.2 kg/m²) and regular sleep/wake habits were studied. Subjects with a personal history of psychiatric illness, endocrine illness, sleep disorders, smoking, drug use, night work, or transmeridian travel in the previous 6 wk were excluded. All subjects provided informed written consent, and all procedures were approved by the Institutional Review Board of the University of Chicago.

Protocol. Four groups of subjects were recruited to participate in studies involving either no nap (control group), a morning nap, an afternoon nap, or an evening nap. Each subject participated in a 4-day study in the Clinical Research Center (CRC) of the University of Chicago. Subjects were requested to keep regular (within 1 h of habitual) bedtimes and wake times for 1 wk before the study. Compliance was verified by analysis of continuous wrist activity recordings (Actiwatch activity monitors with data analyzed using Rhythmwatch software; Mini-Mitter, Sunriver, OR). A schematic representation of the protocol is depicted in Fig. 1. Subjects were admitted to the CRC by 1800 on day 0 for habituation purposes. Bedtime hours were 0000-0800. Sleep recording electrodes were attached, but sleep was not recorded. Dim, reflected, ordinary fluorescent indoor light averaging 36 ± 11 lx (SD) was maintained throughout all waking periods for the remainder of the study. On day 1, subjects had breakfast between 0830 and 0900. An intravenous glucose infusion at a constant rate of 5 g · kg¹ · 24 h¹ was begun at 1000 via a catheter inserted into the antecubital vein in the dominant arm. This glucose infusion constituted the only source of caloric intake until the end of the study on day 4. The subjects had access to water and sugar-free decaffeinated sodas. A second catheter was then inserted in the nondominant arm for blood samples to be drawn at 20-min intervals beginning at 1600 on day 1 and continuing until 0800 on day 4. From 1000 on day 1 until the end of the study on day 4, the subjects remained recumbent in bed with the head of the bed at a 45° angle and continuously awake in constant dim light, except during the scheduled dark/sleep periods. They were allowed to read, work on a computer, watch television, use the phone, have visitors, and play board games. Estimations of baseline (i.e., prestimulus) circadian phase were derived from measurements of plasma melatonin and thyrotropin (TSH) levels during the evening and first part of the night of day 1 obtained under these "constant routine" conditions (i.e., constant dim light exposure and recumbence, constant caloric infusion, and constant wakefulness) as described below. Bedtime hours in total darkness were 0200-0800. Sleep was polygraphically recorded. The delayed bedtime limited the burden of sleep deprivation that would have been imposed by the maintenance of all night constant routine conditions while allowing for the onsets of nocturnal melatonin and TSH secretion to be observed in the absence of masking effects of sleep. The curtailment of the nighttime bedtime period to 6 h also served to facilitate daytime sleep. During day 2, subjects were exposed to a single 6-h period of full recumbence in total darkness, during which they were encouraged to sleep, beginning in the morning (morning nap: 0900-1500; n = 6), the afternoon (afternoon nap: 1400-2000; n = 6), or the evening (evening nap: 1900-0100; n = 6). Polygraphic sleep recordings were performed during all naps. A control group (n = 7) remained continuously awake in constant dim light. After the naps, constant conditions (with enforced wakefulness) resumed throughout day 3 and until 0400 on the morning of day 4, at which time subjects were allowed to sleep as long as they wished. Subjects were released later in the morning after their glucose infusions had been tapered and they had eaten a meal.

View larger version (19K): [in this window] [in a new window]   Fig. 1.   Schematic representation of protocol. Filled rectangles indicate scheduled sleep periods in darkness. Three different groups of subjects were exposed during daytime of experimental day 2 to a single 6-h period of darkness, during which they were encouraged to sleep. A control group ("no nap") remained awake. Blood sampling at 20-min intervals began at 1600 on day 1 of the study (arrow) and continued until 0800 on day 4 (arrow). Plasma profiles of melatonin and thyrotropin (TSH) were used to assess circadian phase before and after stimulus exposure to determine phase-shifting effects of daytime naps in darkness.

Sleep recording and analysis. Polygraphic sleep recordings were visually scored at 30-s intervals in stages wake, I, II, III, IV, and rapid eye movement (REM) using standardized criteria (14) by an experienced scorer. Sleep onset and sleep offset were defined, respectively, as the times of occurrence of the first and last 30-s intervals scored II, III, IV, or REM. Total sleep time was defined as the sum of stages I, II, III, IV, and REM. Sleep efficiency was calculated as the total recording time (i.e., 6 h) minus the total duration of awakenings, expressed in percent of the total recording time. Non-REM (NREM) stages were defined as the sum of stages I, II, III, and IV. Slow-wave (SW) sleep was defined as the sum of stages III and IV.

Assays. Plasma TSH levels were measured by a chemiluminescent enzyme immunometric assay (Immulite Third Generation TSH, Diagnostic Products, Los Angeles, CA). The sensitivity of the assay was 0.002 µU/ml. The intra-assay coefficient of variation averaged 6.2% at 1.3 µU/ml and 3.9% at 3.8 µU/ml. The interassay coefficient of variation averaged 9.5% in the physiological range.

Plasma melatonin levels were measured with a double antibody RIA using commercially available reagents (Stockgrand, Guilford, Surrey, UK) as previously described (18). The lower limit of sensitivity of the assay was 2.5 pg/ml. The intra-assay coefficient of variation averaged 17.5% for values <10 pg/ml, 8.6% in the range of 10-30 pg/ml, and 5.2% for values >30 pg/ml. The interassay coefficient of variation averaged 20% for values <10 pg/ml and 13.5% for values >10 pg/ml.

For all hormonal determinations, samples from the same subject were measured in the same assay.

Estimations of circadian phase. The timings of the onsets of the nocturnal elevations of plasma melatonin and TSH were used to estimate circadian phase before and after stimulus exposure. Phase shifts were defined as the difference between the baseline (i.e., prestimulus) phase estimations on day 1 and the poststimulus phase estimations on day 2 and/or day 3. By convention, phase advances are expressed as positive numbers, and phase delays are expressed as negative numbers.

In the majority of individual studies, daytime melatonin levels were below the limit of sensitivity of the assay (i.e., 2.5 pg/ml). Therefore, the onset of nocturnal melatonin secretion was estimated as the time of the first plasma level >10 pg/ml not followed by a return to concentrations <10 pg/ml. In one subject in the control group, daytime melatonin levels for all 3 days of the study were consistently in the range of 5-10 pg/ml rather than near the limit of detection of the assay. A "basal" level was therefore calculated from the steadily low concentrations measured during the 1600-2200 time interval and the onset of nocturnal melatonin secretion was then estimated as the timing of the first plasma level exceeding that mean basal level +1 standard deviation and not followed by a return to concentrations below this value. One subject in the afternoon group had very low melatonin levels in the range of 5-7 pg/ml without any detectable circadian variation. Because meaningful secretion onsets and phase shifts could not be calculated in this individual, all melatonin data from this subject were excluded from the analysis.

To derive estimations of circadian phase from the profiles of plasma TSH, a best-fit curve using a robust curve-fitting algorithm with a 4-h smoothing window was calculated (3) to define on each study day (i.e., days 1, 2, and 3) the daytime nadir and nocturnal zenith independently of sporadic pulsatile TSH fluctuations. On each study day, the onset of the nocturnal elevation of plasma TSH was then defined as the timing of the first TSH concentration exceeding the midpoint between daytime nadir and nocturnal zenith. In the group exposed to an evening nap (1900-0100) on day 2, the onset of nocturnal TSH secretion was masked by the well-known acute inhibitory effects of sleep and, therefore, TSH-based phase estimations were calculated only for days 1 and 3.

For one subject in the evening group, a technical failure during the assay procedure precluded the determination of sufficient TSH values for an analysis of TSH-derived circadian phase. Therefore, only melatonin-derived phase was available for this subject. Finally, one control subject dropped out on the final night of the study (i.e., day 3-4) for personal reasons, and thus circadian phase assessments could be performed only for days 1 and 2. In summary, for both melatonin and TSH-derived circadian phase, seven control subjects were available for phase shift comparisons from day 1 to day 2, and six control subjects were available for phase shift comparisons from day 1 to day 3. For melatonin-derived circadian phase, comparisons of control were made versus morning (n = 6), afternoon (n = 5), and evening (n = 6). For TSH-derived circadian phase, comparisons of control were made versus morning (n = 6), afternoon (n = 5), and evening (n = 5).

Statistical tests. Comparisons of phase shifts from day 1 to day 2 between the control group and the groups with morning, afternoon, and evening naps were performed using the nonparametric Kruskall-Wallis procedure. When the Kruskal-Wallis H test statistic indicated the presence of between-group differences with P < 0.05, the phase shifts observed in each of the three treatment groups were compared with the phase shifts observed in the control group using a post hoc Dunn's test corrected for ties (22). These calculations were repeated for phase shifts from day 1 to day 3 and for phase shifts from day 2 to day 3 and were performed similarly for both melatonin-based and TSH-based phase estimations.

Unless otherwise indicated, all group data are expressed as means ± SE.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Timing of TSH and melatonin secretion onsets. Under baseline (i.e., prestimulus) conditions, the observed wave shapes of both the plasma melatonin (Fig. 2) and the plasma TSH profiles (Fig. 3) conformed with previous descriptions and showed excellent intersubject and interstudy reproducibility. On average, the melatonin onset on day 1 occurred at 2343 ± 30 min in the control group, 2353 ± 13 min in the morning group, 2248 ± 48 min for in afternoon group, and 2230 ± 24 min in the evening group. Group differences were not significant (ANOVA, P = 0.20). The timings of the nocturnal TSH onsets on day 1 were also similar in all four groups and averaged 2246 ± 14 min for the control group, 2307 ± 16 min for the morning group, 2236 ± 33 min for the afternoon group, and 2158 ± 21 min for the evening group (ANOVA, P = 0.19). A significant correlation was observed between the timings of the onset of nocturnal TSH secretion and of the onset of nocturnal melatonin secretion in the evenings from which both TSH and melatonin-derived circadian phase estimates were possible (n = 63, r = 0.75, P = 0.0001), adding further support to the notion that melatonin and TSH are reliable markers of circadian phase that maintain a consistent phase relationship to each other.

View larger version (38K): [in this window] [in a new window]   Fig. 2.   Profiles of plasma melatonin over 64-h of continuous sampling at 20-min intervals. Group means ± SE profiles are shown at left and representative individual profiles are shown at right. Filled rectangles on axes indicate scheduled sleep periods in fully recumbent posture and darkness.

View larger version (33K): [in this window] [in a new window]   Fig. 3.   Profiles of plasma TSH over 64-h of continuous sampling at 20-min intervals. Group means ± SE profiles are shown at left, and representative individual profiles are shown at right. Filled rectangles on axes indicated scheduled sleep periods in fully recumbent posture and darkness. For each individual, TSH values are expressed as a proportion of overall mean level. On day 1, mean values of TSH acrophase (±SD) were 2.94 ± 1.61 µU/ml; on day 2, 2.64 ± 1.20; and on day 3, 3.09 ± 1.60.

Phase shifts of melatonin and TSH secretion onsets. Over the interval day 1-day 2, differences in phase shifts across subject groups were significant for both melatonin (P = 0.012) and TSH (P = 0.009). Over the interval day 2-day 3, there were no differences between groups for either melatonin or TSH phase shifts. Over the interval day 1-day 3, group differences in phase shifts were significant for both melatonin (P = 0.046) and TSH (P = 0.006).

Figure 4 illustrates the mean phase shifts of both melatonin and TSH for all groups. The control subjects who were not exposed to a nap showed a significant phase delay for both melatonin and TSH on day 2, but no further phase shift from day 2 to day 3. The phase difference between day 1 and day 3 was significant for both markers. Similar drifts in the delaying direction have been previously observed in subjects studied under constant conditions in the absence of exposure to bright or moderate intensity light (2, 18, 23).

View larger version (18K): [in this window] [in a new window]   Fig. 4.   Mean phase shifts (±SE) of onsets of melatonin (left) and TSH (right) secretion over 3 days of sampling. Phase shifts observed in control subjects are depicted by  and connected by a dashed line. Between day 1 (phase estimation prestimulus) and day 2, separate groups of subjects took 6-h naps in darkness starting in the morning (), afternoon (), or evening (). * Significant differences from control. Phase delays, or shifts to later onset timings, are plotted as negative values, whereas phase advances, or shifts to earlier onset timings, are plotted as positive values.

In the morning group, the phase delay of the melatonin onset (77 ± 10 min) from day 1 to day 2 was significantly greater than that occurring in control subjects (37 ± 13 min; P < 0.001). Similarly, a significantly greater phase delay was observed for the TSH onset in subjects who had a morning nap (107 ± 12 min) compared with those who had no naps (40 ± 12 min; P < 0.001). Between day 1 and day 3, differences between the phase delays of melatonin in the morning group (80 ± 9 min) and in the control group (60 ± 23 min) did not reach significance (P > 0.2). However, phase delays of the TSH onset were significantly greater in the morning group (93 ± 18 min) than in the control group (27 ± 20 min; P < 0.001).

In the afternoon group, the phase shifts from day 1 to day 2 as well as from day 1 to day 3 were not different from those observed in the control group for both the melatonin and the TSH onsets.

In the evening group, phase shifts from day 1 to day 2 could only be calculated for the melatonin onset. Contrasting with control subjects who showed a phase delay from day 1 to day 2, the subjects exposed to an evening nap showed essentially no phase drift (77 ± 10 min vs. 7 ± 15 min; P < 0.07). This pattern was maintained through day 3, with phase shifts from day 1 to day 3 averaging 50 ± 32 min for control subjects but only 20 ± 15 min in subjects who had an evening nap (P < 0.001). Estimations of phase shifts from day 1 to day 3 based on the TSH onset indicated that exposure to an evening nap not only prevented the delay drift observed in control conditions but actually resulted in a phase advance (+44 ± 17 min vs. 27 ± 20 min for control; P < 0.001).

Across all studies, the phase shifts of the melatonin onsets and the phase shifts of the TSH onsets were significantly correlated (n = 57, r = 0.56, P = 0.0001).

Phase response curves to daytime naps. Phase response curves (PRC) to daytime sleep in darkness were generated by plotting the magnitude of the phase shift of the melatonin onset or of the TSH onset versus the timing of the center of the nap relative to the respective estimation of circadian phase on day 1. In Fig. 5, the PRC based on the melatonin onset and the PRC based on the TSH onset are aligned vertically with time 0, corresponding to the mean melatonin onset on day 1 (Fig. 5, top; i.e., 2316) and to the mean TSH onset on day 1 (Fig. 5, middle; i.e., 2259). The overall shape of the two PRCs are in good concordance, although phase shifts of TSH onsets were consistently larger in magnitude than phase shifts of the melatonin onsets.

View larger version (15K): [in this window] [in a new window]   Fig. 5.   Partial phase response curves (PRCs) observed in response to daytime sleep in darkness in present study (top) and a schematic PRC to light derived from a previous study (bottom; Ref. 18). By convention, phase advances and phase delays are defined as positive or negative numbers, respectively. Magnitude of phase shift of circadian melatonin onset (top) and TSH onset (middle) for each individual is plotted versus timing of midpoint of 6-h sleep/dark period relative to onset of hormonal phase marker used. Local clock time at zero hour at top also represents mean timing of melatonin or TSH onsets. Zero-hour clocktimes at top are thus aligned with approximate clock time of schematic PRC to light pulses. Timing of daytime naps in darkness is shown by open bars. For visual reference, dashed vertical lines extend from clocktime of morning and evening nap midpoints.

Correlation of sleep with phase shifts. Figure 6 summarizes the characteristics of sleep in the three treatment groups. On average, the subjects slept ~4 h of the scheduled 6-h period of full recumbence in total darkness. Total sleep time and sleep efficiency tended to be higher in the evening than in the morning or afternoon, but differences were not statistically significant. The amount of SW sleep increased significantly from morning to evening (P < 0.002). To address the possibility that the observed phase shifts were related to the amount and/or quality of sleep permitted during the 6-h nap time, correlations were calculated between magnitude of phase shifts (based on both melatonin or TSH onsets) and sleep parameters (latency, efficiency, total sleep time, total minutes of NREM sleep, total minutes of SW sleep, total minutes of REM sleep). The only significant correlation that emerged from this analysis related the amount of SW sleep with the magnitude of the phase shift from day 1 to day 3 (Spearman rank correlation; melatonin onsets: r = 0.81, P < 0.003; TSH onsets: r = 0.79, P < 0.005). Because the morning naps included minimal amounts of SW sleep, if any (Fig. 6), these correlations could simply reflect the association between the largest phase delays with the near absence of SW sleep. However, when only evening naps were examined, a trend toward a similar correlation between magnitude of the phase shift and amount of SW sleep was also detected (melatonin onsets: r = 0.60, P = 0.18, n = 6; TSH onsets: r = 0.90, P = 0.07, n = 5).

View larger version (16K): [in this window] [in a new window]   Fig. 6.   Sleep parameters during daytime naps in darkness. Total sleep time, total sleep of stages 3 and 4, total stage rapid eye movement (REM) sleep, and sleep efficiency (+SE) are shown for subjects taking 6-h daytime naps in darkness beginning in the morning (0900), afternoon (1400), or evening (1900).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

With the use of two independent hormonal markers of circadian phase, the present study demonstrates that daytime naps in darkness presented over a background of very dim light can phase shift human circadian rhythms. A 6-h nap initiated in the morning resulted in rapid delay shifts of both melatonin and TSH onsets, which were, on average, 40 min greater than those observed in control subjects who remained continuously awake in dim light. A 6-h afternoon nap resulted in no consistent phase changes in either phase marker on the same day or on the following day. Subjects exposed to evening naps experienced phase advances of both melatonin and TSH onsets relative to control subjects. The phase shifts in response to exposure to daytime dark periods observed in the present study thus conform with the classical concept of a phase response curve (5).

The magnitudes of the shifts of TSH onsets were generally larger than that of the shifts of melatonin onsets. However, the phase shifts derived from the melatonin onset were in the same relative direction (i.e., either advanced or delayed relative to control) and quantitatively correlated with those derived from the TSH onset. Because the neural pathways subserving the circadian control of pineal function on the one hand and of hypothalamic-pituitary-thyroid axis function, on the other hand, are clearly different (21), such discrepancies are not unexpected.

Although the present study represents the first attempt to systematically characterize the response of the human circadian system to sleep/dark pulses, alterations of circadian phase in response to a shifted sleep/dark period have been described in two recent reports (11, 17). In one study, an 8-h advance of the sleep-wake cycle (i.e., an 8-h "nap" centered at 1900) was associated with a nearly 2-h phase advance of the melatonin onset. The background light intensity during waking periods was ~200 lx. Although we did not observe absolute phase advances in response to evening naps using the melatonin onset as a circadian marker in the present study, the TSH onset was significantly phase advanced and, for both markers, the subjects were significantly phase advanced relative to controls. This suggests that a human PRC to sleep/dark pulses using melatonin onset as a phase marker may include a more clearly defined phase advance region than was observed in the present study if the dark pulses are presented over a background of light intensity brighter than the 36 lx used in the current study. A second recent study examining circadian phase in response to an 8-h delay shift of the sleep/dark period (i.e., an 8-h "nap" centered at 1100), with background light of <100 lx during wake periods, revealed consistent 2-3 h phase delays of markers of circadian phase derived from melatonin, TSH, and cortisol profiles (11). In contrast to these two reports of phase shifts in response to daytime sleep in darkness, in another study of the phase-shifting effects of a sleep/dark period shifted by 12 h, with background light of <15 lx in intensity, only a slight drift of the rhythm of core body temperature was observed (6). However, the timing of the midpoint of this daytime sleep/dark period is roughly equivalent to the timing of the afternoon naps in the present study, which also did not result in significant phase shifts. The findings in the present study are thus remarkably consistent with these previous data and suggest that the magnitude of the advance and delay regions in the human PRC to daytime naps in darkness may depend on the magnitude of the contrast between the light and the dark.

Although the majority of studies of human entrainment has focused on the timing and intensity of light pulses, the present data and the two studies (11, 17) that have examined 8-h shifts of the sleep/dark period provide strong evidence for a role of exposure to sleep/dark on circadian phase. Further understanding of the patterns of entrainment of human rhythms will thus require considering the temporal pattern and contrast of the overall light-dark cycle as well as of the timings of the light-dark and dark-light transitions rather than focusing only on the timing, duration, and intensity of light exposure (7). Indeed, phase shifts thought to represent the effects of light alone using experimental designs, including exposure to both light and dark (4), may actually reflect the combined effects of light and dark.

The phase-shifting effects of daytime sleep in darkness could reflect nonphotic effects of sleep rather than effects of dark per se. In nocturnal rodents, there is good evidence to indicate that changes in the rest-activity state may phase shift free-running circadian rhythms or modify the pattern of entrainment to the light/dark cycle. Exposure to stimuli that result in an increase in the level of activity during the usual rest period is associated with robust phase shifts, thus providing evidence for the existence of a nonphotic input pathway to the circadian system (15, 16). Conversely, there is some evidence from hamster studies that enforced immobilization during the beginning of the usual active period results in phase delays of the rhythm of locomotor activity (19). The phase delays observed in the subjects exposed to morning naps in the present study could reflect a similar phenomenon. Sleep was polygraphically recorded during all naps to examine whether the direction and magnitude of phase shifts were related to parameters quantifying the duration, fragmentation, and quality of sleep. The only significant correlations reflected the expected coincidence of the phase delays observed in response to the morning nap with the near total absence of SW sleep and of the phase advances recorded after the evening nap with the largest amounts of SW sleep. This association may not involve common mechanisms. However, the possibility that processes involved in SW sleep regulation may affect circadian function should not be excluded because the correlation between the amount of SW sleep and the magnitude of the phase shift remained apparent when only the evening group was considered. Further studies will be needed to address the possible role of sleep in regulation of circadian function.

Alternatively, the phase shifts observed in the present study could reflect photic effects of the dark-light contrast independent of the wake or sleep state. Data in support of phase-shifting effects of dark per se, independently of the sleep/rest or wake/activity state, were recently reported in the diurnal rodent Octodon degus. Indeed, in this species, phase shifts to dark pulses were apparently not related to a reversal of the rest/activity levels because no acute activity decreases reflecting a possible induction of sleep were observed (13). The present findings may indicate that exposure to daytime periods of darkness similarly modulates human circadian phase. The mirrorlike relationship of the PRCs derived from the present data with a schematic representation of a single-pulse PRC to dim light (Fig. 5) suggests that dark exposure may actually modify circadian phase by preventing the effects of dim light. The response to dark (i.e., the absence of light) in the present study resulted in phase delays during the early subjective day and relative phase advances in the late subjective day. This pattern is a mirror image of the light PRC and suggests that dark pulses may phase shift circadian rhythms by altering photic synchronization. Further controlled studies are needed to resolve the differential and combined effects of darkness and sleep on circadian function.

Appropriately timed exposure to both light and periods of daytime dark/sleep could be used to facilitate shifts to a new pattern of entrainment or to stabilize entrainment to a beneficial or useful schedule. The importance of periods of darkness and the timing of the absence of light for the beneficial modulation of circadian timing and entrainment is supported by field studies in which the phase-shifting effects of appropriately timed bright light to reentrain night workers were obtained only with the consistent use of appropriately timed darkness. The use of black sheeting on the bedroom windows provided a sharply contrasting light/dark cycle to facilitate induction of phase shifts and reentrainment in a real world environment (8). The use of dark welders' goggles was used to reduce the confounding effects of light during the transit home from work (9).

On the other hand, inappropriately timed exposure to dark/sleep may result in circadian misalignment. Once entrained to a particular light and dark schedule, the customary practice of sleeping later on the weekends, although satisfying a sleep need induced by cumulative sleep loss during the week, may actually delay the circadian clock and hinder adaptation relative to work and social demands. The absence of a phase-shifting effect of afternoon naps in the present study suggests that an afternoon siesta is appropriately timed to alleviate sleep need without altering circadian phase.

In conclusion, using two independent hormonal markers of circadian phase, we have demonstrated in the present study that daytime naps in darkness can impact the phase of human circadian rhythms. The overall contrast and temporal pattern of exposure to both light and dark/sleep appear therefore to convey temporal cues to the human circadian pacemaker.

Perspectives

The results of the present study indicate that the timing of sleep and/or exposure to darkness can phase shift human endocrine rhythms, consistent with two other recent reports (11, 18). These results may be particularly relevant to those engaging in shift work or recovering from sleep deprivation with daytime naps. Both dark and sleep may have played a role in causing the phase shifts observed in the present study. Other studies are needed to determine their respective contributions. Experiments that vary only illumination levels during continued wakefulness will be needed to determine whether the observed phase shifts are due to dark/light alone. If the observed phase shifts are due to nonphotic mechanisms alone (i.e., sleep), then the present study suggests that sleep-wake states play a modulatory role in circadian timing. Experiments that manipulate sleep quality (i.e., amount and proportions of sleep stages) will be required to delineate the mechanisms underlying modulation of circadian timing by sleep.

When coupled with the well-known phase-shifting effects of light on human circadian rhythms, our results suggest that the use of combined strategies involving both photic and nonphotic cues may enable individuals to more rapidly adjust their biological clock to imposed schedules often encountered in modern life.