Homeostatic Regulation of REM Sleep in Humans During Extended Sleep

Homeostatic Regulation of REM Sleep in Humans During Extended Sleep Giuseppe Barbato and Thomas A. Wehr Clinical Psychobiology Branch, National Institute of Mental Health, Bethesda Md Summary: Benington and Heller (1994) recently proposed a sleep-dependent model for the homeostatic control of REM sleep in which the amount of REM sleep propensity discharged in each bout of REM sleep affects the timing of the subsequent REM episode. Consistent with their hypothesis, they reported that in rats the duration of a REM episode was positively correlated with the duration of the succeeding NREM episode and not with the duration of the preceding NREM episode. To assess this hypothesis in humans, we used 308 sleep records from 11 subjects who remained at bedrest in the dark and slept ad libitum during 14-hour periods each night for 4 weeks. The timing of the onset of the first REM episode of the long night was linked to the timing of sleep onset. NREM-REM cycle duration decreased progressively throughout the night as a result of a progressive decrease in duration of the NREM component. Durations of REM sleep episodes correlated significantly with durations of subsequent NREM episodes in three out of the eight rank cycles analyzed (p<.0031, Bonferroni corrected); positive correlation coefficients were found for all the remaining cycles, but were not statistically significant when the conservative Bonferroni correction of the alpha level was applied. With the exception of the first sleep cycle, durations of REM sleep episodes did not correlate with durations of preceding NREM sleep episodes. According to the present results, the amount of REM sleep in one episode controls the time of occurrence of the next REM episode when the impact of other possible regulating factors are at a low level. We hypothesize that the extended dark/rest period, by increasing the time window allowed for sleep, provided a condition under which the systems governing REM sleep expression were free of the masking imposed by the conventional 16 hours light/8 hours dark schedule that consolidates and compresses sleep. Key words: NREM sleep; REM sleep; NREM-REM cycle, extended sleep TWO MAIN THEORIES have been proposed to explain the nature of the process that governs the occurrence of rapid eye movement (REM) sleep. According to the oscillatory hypothesis, 1 REM sleep is an expression during sleep of the rhythm of an oscillator, the basic rest-activity cycle (BRAC), that operates continously during both wakefulness and sleep. According to the sleep-dependent hypothesis, 2 each bout of REM sleep is triggered by the occurrence of the bout of non-rem (NREM) sleep that precedes it. Accepted for publication December, 1997 Address correspondence and requests for reprints to Dr. Giuseppe Barbato Dipartimento di Neuroscienze e Comunicazione Interumana, Sezione di Psichiatria, Via Pansini 5, 80131, Napoli - ITALY The two models imply different types of correlations between durations of successive NREM and REM periods. 3 The oscillatory model would predict a negative correlation to maintain a constant cycle length. The sleep-dependent model would predict a positive correlation to satisfy homeostatic rules. Benington and Heller 4 have recently proposed a sleepdependent model for the homeostatic control of REM sleep in which the amount of REM sleep propensity discharged in each bout of REM sleep affects the timing of the subsequent REM episode. If less REM propensity is discharged, then a shorter interval of NREM sleep will intervene before the subsequent REM sleep episode. On the other hand, longer REM episodes discharge more REM propensity, so that a longer NREM interval passes before the occurrence SLEEP, Vol. 21, No. 3, 1998 267

of the next REM episode. Consistent with this hypothesis, they found that the duration of REM sleep episodes in rats were positively correlated with the duration of subsequent NREM sleep episodes. A limitation of this type of analysis in humans is that normally only four or five NREM- REM cycles occur each night. A previously published study 5 in which we had allowed individuals to sleep ad libitum for 14 hours every night for 4 weeks seemed particularly well-suited to test Benington and Heller's hypothesis in humans, because it provided us with a data base with a high number of sleep cycles per night and a large number of nights per individual. METHOD The analysis was carried out on 308 sleep records of 11 healthy male volunteers, age 20-34 years, who participated in an experiment that was designed to investigate the effects of alterations in duration of the photoperiod (the illuminated fraction of the day) on sleep and endocrine systems. 5 The subjects lived for 4 weeks in a winter-type photoperiod in which they went about their usual activities in ambient natural and artificial light for 10 hours (08:00-18:00) each day and were confined to a dark room for 14 hours (18:00-08:00) each night. The subjects were instructed to remain at bedrest and to sleep whenever possible during the dark period. Polysomnographic sleep was monitored with a Grass 78 D polygraph. The sleep stages were visually scored according to the criteria of Rechtschaffen and Kales. 6 Sleep onset was defined as the first minute of 3 consecutive minutes of stage 2, 3,4 or REM. Since sleep was more fragmented than usual in these conditions, owing to the more frequent occurrence of spontaneous awakenings, restrictive criteria for the definition of a REM period and a NREM-REM cycle were needed. A conservative 3-minute rule 7 was used to differentiate REM periods, according to which sequences of REM sleep were combined into one REM episode if the interval between them was less than 3 minutes. In agreement with previous studies, 8,9 a complete NREM-REM cycle was defined by the successive occurrence of a NREM episode and of a REM episode. The end of the cycle was defined by the occurrence of a new NREM episode or by the occurrence of wakefulness. Incomplete cycles, starting with a NREM episode but not followed by a REM episode (missed REM periods), and REM episodes not preceded by NREM episodes (REM-onset periods), were not included in the analysis of the NREM-REM cycles. No criterion for minimum duration of a NREM- REM cycle was adopted, to guarantee the inclusion in the analysis of even the shortest cycles that occurred in the extended sleep paradigm. Figure 1. Frequency distribution of sleep latency, REM latency, and REM latency from lights-off. Pooled data relative to the first eight cycles for the eleven subjects were analyzed using non-parametric Spearman rank test, corrected for ties correlation coefficients (rs). A Bonferroni correction of the alpha level for SLEEP, Vol. 21, No. 3, 1998 268

included only the NREM periods that fell between two successive REM periods. REM onset periods were also included in this analysis. An exploratory analysis of patterns across the night of the NREM-REM cycles, NREM episodes and REM episodes was performed on the averages of the nine subjects who had available data for the eight rank cycles. A post-hoc analysis was conducted on REM and NREM episodes that followed a sleep interruption. In order to be consistent with previous studies on the effect of awakenings on sleep cycles, 10,11,12 this analysis was conducted on cycles following at least 10 minutes of wakefulness (stage 0). A sleep-onset REM period (SOREM) was defined as one having a REM latency shorter than 25 minutes. RESULTS Figure 2. Average profiles of NREM-REM cycles, NREM periods and REM periods. Analyses are on th4 eaverage of the nine subjects who had available data for the eight rank cycles. Average times of day for each cycle were: 21:46 ± 47 minutes for cycle 1, 23:21 ± 43 minutes for cycle 2, 00:40 ± 61 minutes for cycle 3, 02:42 ± 44 minutes for cycle 4, 04:14 minutes for cycle 5, 05:13 ± 39 minutes for cycle 6, 05:50 ± 39 minutes for cycle 7, 06:08 ± 38 minutes for cycle 8. the number of analyses (n=16) was adopted: alpha =0.5/16=0.0031. Analysis of the correlation between the duration of NREM periods and the duration of following REM periods included all complete NREM-REM cycles. Analysis of the correlation between the duration of REM periods and the duration of following NREM periods SLEEP, Vol. 21, No. 3, 1998 269 A total of 1676 complete NREM-REM cycles were available for the analysis. Tables 1, 2 and 3 show the data for duration of NREM-REM cycles and for their NREM and REM components for each of the eleven subjects. Mean sleep latency was 140.2 ± 67.0 min. (Fig. 1), mean REM latency was 77.7 ± 32.4 minutes (Fig. 1), and mean REM latency from lights off (18:00) was 218.4 ± 71.3 minutes(fig. 1). No statistically significant relationships between sleep latency and either REM latency or first cycle REM duration were found. All subjects showed statistically significant differences of NREM-REM cycle durations across the night (Table 1). Analysis for each rank cycle also showed significant differences between subjects (Table 1). Average profiles of NREM-REM cycles, NREM periods and REM periods are shown in Fig. 2. NREM-REM duration decreased progressively throughout the night, mainly as a result of a progressive decrease of the NREM component. Variation of the NREM-REM cycle duration across the night was best fitted by a linear regression (y= 102.002-5.658x), as was the case for the NREM component (y = 82.597-6.134x). REM duration showed low values in the first and last cycle, leading to a curvilinear profile, which was best fitted by a quadratic regression (y = 7.585 + 7.567x -.788x 2 ). Analysis of the relationships between REM episode duration and previous (Fig. 3) and successive NREM episode durations (Fig. 4), revealed a statistically significant relationship between REM duration and previous NREM duration only for the first cycle (Table 4). In contrast, a statistically significant positive correlation between REM duration and successive NREM duration was observed for three of the cycles analyzed (Table 4). Two hundred twenty-five REM periods followed wakefulness interruptions of at least 10 minutes. Mean REM latency for the REM periods was of 32.2±24.3 min-

Figure 3. Scatter plot of duration of NREM periods and duration of following REM periods SLEEP, Vol. 21, No. 3, 1998 270

Table 1. Complete(a) NREM-REM cycle duration (minutes). 11 subjects cycle 1 cycle 2 cycle 3 cycle 4 cycle 5 cycle 6 cycle 7 cycle 8 Subject N mean± s d N mean± s d N mean± s d N mean± s d N mean± s d N mean± s d N mean± s d N mean± s d A 26 87.6±39.3 21 100.5±29.1 25 94.4±28.7 27 85.2±22.5 20 77.4±21.8 18 47.1±22.5 7 59.8±28.7 3 70.5±19.2 F=6.8 * B 23 103.9±30.5 20 110.1±35.3 24 81.9±29.7 27 79.6±23.4 23 59.9±32.0 21 52.1±34.2 14 42.4±21.3 4 38.1±18.7 F=13.2 * C 20 87.3±27.5 24 80.8±23.9 25 79.7±14.9 23 81.8±16.1 21 75.3±23.1 24 68.9±29.8 20 56.3±13.4 12 66.0±24.3 F=4.1 ** D 20 82.1±24.0 20 94.9±21.9 16 79.4±38.2 18 68.9±32.9 17 70.5±25.1 18 70.0±24.0 15 67.2±21.3 14 60.6±36.1 F=2.6 **** E 27 116.7±64.0 26 86.0±21.5 25 80.4±17.5 26 73.0±28.2 27 88.4±17.3 22 70.7±29.9 17 61.3±27.1 11 58.9±27.4 F=6.7 * F 20 69.3±22.3 20 94.9±35 16 76.3±31.0 19 76.7±33.8 18 60.4±34.6 15 65.5±32.8 11 59.7±27.2 13 47.1±28.9 F=3.5 *** G 23 76.9±13.8 21 98.4±39.0 26 97.2±31.3 22 100.0±27.4 20 90.9±41.4 20 75.3±30.2 18 65.1±32.6 9 52.9±28.9 F=4.8 * H 28 94.0±25.6 23 130.1±34.6 26 101.9±41.6 28 94.0±27.8 25 78.8±36.4 12 69.8±31.0 4 39.9±26.5 F=8.5 * I 28 90.8±48.6 28 100.0±15.6 26 101.9±25.4 26 101.1±45.0 26 75.5±26.5 22 75.3±19.8 10 59.3±26.0 3 64.0±13.8 F=4.2 * L 28 109.5±32.2 27 92.5±16.8 27 91.2±23.5 27 78.6±18.0 26 60.8±26.7 7 47.0±22.8 3 31.9±29.0 F=15.4 * M 25 88.1±22.2 24 92.4±20.1 19 86.0±30.6 18 80.1±14.7 14 87.8±25.7 16 72.4±19.9 9 64.4±21.3 9 49.5±17.5 F=5.2 * F= 3.72 * F= 5.4 * F= 2.5 # F= 3.4 ** F= 3.3 ** F= 2.3 **** F= 1.4 F= 1 *p=.0001 ** p=.0005 # p<.001 *** p=.002 # p=.001 **** p=.02 (a): NREM episodes followed by a REM episode, incomplete cycles are not computed in the analysis Table 2. NREM period duration (minutes) of complete(a) NREM-REM cycles. 11 subjects cycle 1 cycle 2 cycle 3 cycle 4 cycle 5 cycle 6 cycle 7 cycle 8 Subject N mean± sd N mean±sd N mean±sd N mean±sd N mean±sd N mean±sd N mean±sd N mean± s d A 26 72.8±29.8 21 70.8±21.9 25 62.3±23.7 27 55.1±16.3 20 40.2±17.2 18 23.9±15.2 7 36.5±27.2 3 49.3±20.6 F=11.56 * B 23 86.0±27.2 20 90.8±33.8 24 60.8±26.6 27 57.6±19.3 23 34.9±19.8 21 31.8±19.7 14 28.5±18.0 4 25.3±17.8 F=20.34 * C 20 70.1±26.8 24 62.4±20.5 25 65.2±16.2 23 60.1±14.8 21 52.5±19.5 24 47.5±24.1 20 37.4±11.0 12 44.2±18.7 F=6.72 * D 20 67.9±19.3 20 80±19.7 16 57.6±28.0 18 49.2±28.0 17 45.3±21.0 18 45.9±21.5 15 47.9±16.8 14 36.6±22.2 F=7.19 * E 27 102.6±55.1 26 70.6±20.0 25 59.7±16.7 26 48.9±26.4 27 58.1±19.7 22 44.6±28.8 17 38.9±24.1 11 36.8±24.6 F=11.70 * F 20 57.1±17.9 20 73.8±34.1 16 51.2±29.9 19 51.7±28.3 18 35.7±25.3 15 39.7±25.4 11 33.8±28.4 13 26.4±26.0 F=5.30* G 23 66.3±11.3 21 80.6±35.3 26 79.8±29.0 22 75.8±30.6 20 63.9±32.0 20 43.1±19.9 18 43.1±23.6 9 40.6±26.5 F=7.20 * H 28 74.1±20.6 23 103.7±31.3 26 74.8±39.1 28 60.0±27.5 25 48.9±27.1 12 46.5±26.5 4 29.0±26.3 F=10.50 * I 28 77.5±40.5 28 77.3±13.5 26 76.4±21.5 26 69.9±37.9 26 48.6±22.1 22 52.6±17.6 10 40.1±16.1 3 55.5±20.3 F=5.70 * L 28 96.6±30.8 27 72.5±10.9 27 67.2±15.8 27 55.0±15.6 26 35.6±20.6 7 28.4±16.0 3 15.5±19.6 F=29.98 * M 25 72.4±18.5 24 71.6±17.2 19 63.2±18.9 18 58.2±12.0 14 61.1±16.0 16 45.9±12.7 9 38.4±13.0 9 31.9±9.8 F=12.40 * F= 4.8 * F= 5.0 * F= 2.7 ** F= 2.6 ** F= 4.5 * F= 2.9 ** F= 1.1 F= 1.3 *p=.0001 ** p=.005 (a): NREM episodes followed by a REM episode, incomplete cycles are not computed in the analysis utes. Of these periods, 110 (40%) were further characterized by a REM latency shorter than 25 minutes. DISCUSSION The duration of the NREM-REM cycle does not appear to be as stable as one would predict if it were driven by a regular rhythmic process. 13,14 Several authors have reported large variability in cycle SLEEP, Vol. 21, No. 3, 1998 271 duration, 15-18 and others have reported a systematic trend in cycle length across the sleep period in grouped data that would not be apparent in a single sleep period. 8 Consistent with these earlier observations, we found a high variability in the duration of the NREM-REM cycle in all our subjects, and a progressive shortening of the average duration of the cycle across the night due to a shortening of the duration of its NREM component. A similar progressive decrease in NREM period dura-

Table 3. REM period duration (minutes) of complete(a) NREM-REM cycles. 11 subjects cycle 1 cycle 2 cycle 3 cycle 4 cycle 5 cycle 6 cycle 7 cycle 8 Subject N mean± s d N mean± sd N mean± sd N mean± sd N mean± s d N mean±sd N mean± sd N mean± sd A 26 14.8±10.5 21 29.7±13.4 25 32.2±16.0 27 30.0±14.9 20 37.1±20.3 18 23.1±14.8 7 23.3±9.3 3 21.2±3.8 F=4.88 * B 23 17.9±9.8 20 19.3±10.8 24 21.1±14.7 27 22.0±14.9 23 24.9±21.8 21 20.3±23.0 14 13.9±11.3 4 12.9±6.2 F=0.85 C 20 17.2±8.4 24 18.4±7.5 25 14.6±8.2 23 21.7±11.9 21 22.8±13.7 24 21.4±19.8 20 18.9±11.5 12 21.8±7.5 F=1.23 D 20 14.3±8.1 20 14.9±8.1 16 21.8±21.6 18 19.7±13.1 17 25.2±13.6 18 24.2±16.0 15 19.4±11.4 14 23.9±19.8 F=1.57 E 27 14.1±10.4 26 15.4±7.3 25 20.7±9.0 26 24.1±16.3 27 30.3±14.7 22 26.1±16.1 17 22.4±12.0 11 22.1±12.9 F=4.62 * F 20 12.2±6.1 20 21.1±15.4 16 25.1±19.1 19 24.9±14.6 18 24.6±16.2 15 25.9±21.2 11 26.0±16.1 13 20.7±16.3 F=1.6 G 23 10.6±5.3 21 17.8±9.2 26 17.3±6.9 22 24.2±16.1 20 27.0±16.2 20 32.1±21.3 18 22.0±19.9 9 12.3±6.5 F=5.2 * H 28 19.9±9.1 23 26.3±13.6 26 27.2±13.5 28 34.0±12.3 25 29.9±17.7 12 23.3±10.0 4 10.9±2.4 F=4.1 ** I 28 13.3±9.5 28 22.8±8.1 26 25.4±9.7 26 31.1±14.8 26 26.9±12.5 22 22.8±9.8 10 19.3±18.2 3 8.5±6.9 F=6.2 * L 28 12.9±6.4 27 20.1±11.6 27 24.0±13.7 27 23.6±11.9 26 25.2±13.4 7 18.9±12.1 3 16.3±10.8 F=3.5 *** M 25 15.8±6.6 24 20.8±11.5 19 22.8±15.9 18 21.9±10.7 14 26.8±17.8 16 26.5±17.7 9 25.9±21.3 9 17.6±9.0 F=1.5 F=2.5 # F=3.7 * F=3.1 ** F= 2.6 # F= 1.2 F= 0.7 F= 0.9 F= 1.0 *p=.0001 ** p=.001 *** p=.005 # p=.01 (a): NREM episodes followed by a REM episode, incomplete cycles are not computed in the analysis Table 4. Correlation between REM sleep episode duration and NREM sleep episode duration NREM/REM REM/next NREM # cycle N rs p* N rs p* 1 268 0.431 0.0001 230 0.125 0.06 2 254 0.11 0.11 229 0.323 0.0001 3 255 0.029 0.64 239 0.174 0.0072 4 261-0.09 0.15 231 0.22 0.0009 5 237 0.072 0.27 182 0.177 0.017 6 188 0.067 0.36 114 0.409 0.0001 7 132-0.01 0.89 61 0.286 0.0265 8 81 0.21 0.032 34 0.37 0.0327 # NREM episodes followed by a REM episode * alpha level of statistical significance: p<.00312 (Bonferroni corrected) SLEEP, Vol. 21, No. 3, 1998 272 tion across the night was previously reported by Feinberg et al 19 and by Dijk et al. 20 In the study of Feinberg et al, NREM period durations showed a large decline between the third and fourth period and a small but also consistent fall between the fourth and fifth period. Little change occurred between the fifth and the seventh period. In the study by Dijk et al, the longest NREM sleep episodes occurred at the beginning of sleep, while the shortest ones occurred in the 7th and 10th of 10 consecutive 1.5-hour intervals into which a 15-hour extended sleep period was divided. According to homeostatic models of sleep regulation, 21,22,23 the decreasing trend of NREM period duration across successive cycles, together with a similar trend for EEG delta, reflects the progressive discharge during sleep of a sleep-promoting process that increases during wakefulness. As this process is discharged, the propensity for NREM sleep and EEG delta sleep decreases, while the propensity for REM sleep and arousal increases. 24 Consistent with previous reports, 25,26,27 the timing of the onset of the first REM episode was not systematically related to the timing of lights-off or clock time, but was linked to the timing of sleep onset, occurring 60-70 minutes later. In addition, the duration of this REM episode was positively correlated with the duration of the previous NREM episode. Several authors 21,25,28,29 have suggested that factors that regulate the expression of NREM and REM components in the first sleep cycle of the night are different from those that regulate their expression in subsequent cycles. The first cycle is the one in which the largest amount of slowwave sleep (SWS) is expressed, 22,30 and the extent to which slow-wave sleep is expressed in the NREM period of this cycle appears to play a major role in determining the timing and the intensity of expression of REM sleep in the following REM period. 31 This relationship appears to be antagonistic, 32,33 so that when slow wave sleep propensity is high (as for example after sleep deprivation), the onset of REM sleep is delayed and its intensity is diminished. As the amount of SWS declines exponentially in NREM periods of subsequent cycles, 9,22 SWS propensity and NREM period duration seem to play a much less important role in the regulation of REM periods of these cycles. Our data from these subsequent cycles are in accordance with the homeostatic hypothesis proposed by Benington and Heller and with their data in rats. Durations of REM sleep periods were positively correlated with the durations of NREM sleep periods that immediately followed them.

Figure 4. Scatter plot of duration of REM periods and duration of the following NREM periods SLEEP, Vol. 21, No. 3, 1998 273

Similar correlations have been reported in monkeys 34,35 and cats. 36 These results also are consistent with an observation of Tissot that the longer a sleep stage lasts, the more time intervenes before it recurs. 37 On the other hand, our data are not in agreement with those of Hartmann, 3 who found a positive correlation in mid-sleep sleep cycles between the length of a NREM sleep period and the length of the subsequent REM period. Differences in subject characteristics (patients versus healthy volunteers), sleep schedules (conventional versus extended sleep), and type of data (single versus pooled subjects), make it difficult to compare the results of the two studies. The positive correlations between the duration of REM sleep episodes and the duration of the following NREM sleep episodes raise a possibility, previously suggested by Feinberg 21 and Horne, 38 that REM sleep plays a role in the regulation of NREM sleep propensity. In a recent version of the two-process model of sleep regulation, Achermann et al 39 propose that process S builds up not only during waking but also at any time during sleep that slow-wave activity (SWA) is low. This refinement of the model could explain the resurgence of slow wave sleep that sometimes occurs at the end of extended sleep when, according to the original version of the homeostatic model, the need for SWA should have been completely discharged. Because of the unusual nature of the extended sleep schedule used in this study, the results should be interpreted cautiously. They might depend, for example, on the fragmentation of sleep that occurred in these conditions, and might not pertain to sleep in other conditions in which this fragmentation did not occur. Also, the data from sleep in these conditions was unusual in that it included a substantial number of episodes of REM sleep that were followed by transitions to episodes of spontaneous wakefulness instead of transitions to episodes of NREM sleep. It is possible that the periods of wakefulness may have partly contributed to the interactions between REM and NREM sleep durations that we observed. In this regard, in a previously published analysis of sleep 7 in eight of the eleven subjects reported here, we found that the REM sleep episodes of cycles that were interrupted by wakefulness were shorter than those that were not interrupted. The frequency of interruptions was higher in the later cycles, consistent with a progressive increase of arousal across the night. Previous studies have shown that episodes of intermittent awakening from sleep have a significant effect on the REM sleep that occurs subsequently. 10,11,40,41 REM latencies are shortened and REM durations are lengthened in the REM periods that follow such awakenings. 42 These changes have been considered to represent a rebound of REM sleep that results from REM sleep suppression that is imposed by the period of wakefulness. According to the model proposed by Benington and Heller, interruption of REM sleep by a period of wakefulness would prevent REM sleep propensity from reaching its lowest levels, so that a shorter amount of NREM sleep would be required to trigger REM sleep in the sleep which follows. Consistent with this interpretation, we found, in a post hoc analysis, that REM latency was reduced following sleep periods that were interrupted by wakefulness, with 40% of cases meeting criteria for SOREMs. Data of Foret et al 11 are also consistent with this interpretation. They found that forced interruption of sleep shortened the inter-rem interval if the awakening occurred during an episode of REM sleep, but that it lengthened the inter-rem interval if it occurred during an episode of NREM sleep. The polyphasic sleep pattern that emerged in our subjects when they slept in extended nights seems to have allowed us to observe during spontaneous transitions to wakefulness the same effects that Foret et al demonstrated with forced awakenings. A previous study in monkeys, 34 who regularly exhibit a polyphasic sleep pattern, has shown that the sleep cycle is driven by REM sleep. Campbell and Zulley 43 have suggested that the expression of putative non-circadian components of the human sleep-wake system may be masked by an externally imposed monophasic sleep regimen. Adults who sleep in temporal isolation or in enforced bed rest experiments, 44,45,46 and infants, 47 who do not yet conform to adult sleep schedules, show a clear tendency to have polyphasic patterns of sleep like those that occur in many animals. Elsewhere, 48 we have speculated that the present experiment, with its 14-hour nightly window for sleep, may provide conditions that permit a more natural pattern of human sleep to be expressed. We might also speculate that these conditions permit systems that regulate REM sleep to follow a more natural physiologic course, free of masking effects of modern sleep schedules that compress and consolidate sleep. The extended sleep schedule also may have lowered NREM sleep propensity 7,49 and thereby reduced its influence on the expression of REM sleep, which is antagonistic, according to Borbely and Achermann. 50 Consistent with Benington and Heller's results in rats, the coefficients of the correlations between durations of REM periods and subsequent NREM periods were generally low, in some cases not surviving the conservative Bonferroni correction for statistical significance, suggesting that homeostatic mechanisms are not the only factors affecting REM sleep duration. Another process that seems to strongly affect the regulation of REM sleep, and which might compete with the homeostatic process that we have examined in our analysis, is a circadian rhythm that regulates REM sleep propensity. Data from different sleep studies have shown that the daily SLEEP, Vol. 21, No. 3, 1998 274

peak in REM sleep propensity tends to occur on the rising slope of the body temperature circadian rhythm in the latter part of the night. 28,51,52,53,54 It could be hypothesized that REM episodes occurring in proximity to the trough of temperature curve are less sensitive to the homeostatic control. A role for circadian processes in REM sleep regulation was also suggested by results of a recent study, 55 in which we found preliminary evidence that REM sleep expression might be facilitated by the secretion of cortisol, the circadian rhythm of which reaches its peak at the end of the nightly sleep period. In conclusion, the present results are in favor of a model of REM sleep regulation which includes more than one factor. Heightened propensity for slow-wave activity in the first NREM period may interfere with the expression of REM sleep. In subsequent cycles, REM sleep appears to regulate itself by means of a homeostatic control. In the last part of the sleep period, the circadian rhythm of REM sleep propensity seems to become the predominant influence upon its regulation. REFERENCES 1. Kleitman N. Sleep and wakefulness. University of Chicago Press, Chicago,1963. 2. Moses J, Lubin A, Johnson BL, Naitoh P. Rapid eye movement cycle is a sleep dependent rhythm. Nature 1977; 265:360-361 3. Hartmann E. Mechanism underlying the sleep-dream cycle. Nature 1966; 212:1118-1119. 4. Benington JH, Heller HC. REM-sleep timing is controlled homeostatically by accumulation of REM sleep propensity in non-rem sleep. Am J Physiol 1994; 266:R1992-R2000. 5. Wehr TA, Moul D, Barbato G, Giesen HA, Seidel JA, Barker C, Bender C. Conservation of photoperiod-responsive mechanisms in humans. Am J Physiol 1993; 265:R846-R855. 6. Rechtschaffen A, Kales A. A manual of standardized terminology techniques and scoring system for sleep stages of human subjects. National Institutes of Health, Publication 204 US Government Printing Office, Washington, DC, 1968. 7. Barbato G, Barker C, Bender C, Giesen HA, Wehr TA. Extended sleep in humans in 14 hour nights (LD 10:14): relationship between REM density and spontaneous awakening. Electroencephalogr Clin Neurophysiol 1994; 90:291-297. 8. Feinberg I, Floyd TC. Systematic trends across the night in human sleep cycles. Psychophysiology 1979; 16:283-291. 9. Dijk DJ, Brunner DP, Borbely AA. Time course of EEG power density during long sleep in humans. Am J Physiology 1990;258:R650- R661. 10. Schultz H, Tetzlaff W. Distribution of REM latencies after sleep interruption in depressive patients and control subjects. Biol Psychiatry 1982; 17:1367-1376. 11. Foret J, Touron M, Clodoré M, Benoit O, Bouard G. Modification of sleep structure by brief forced awakenings at different times of the night. Electroencephalogr Clin Neurophysiol 1990; 75:141-147. 12. Miyasita A, Fukua K, Inugami M. Effects of sleep interruption on REM-NREM cycle in nocturnal human sleep. Electroencephalogr Clin Neurophysiol 1989; 73: 107-116. 13. Webb WB, Dube MG. Temporal characteristics of sleep. In: Aschoff J, ed. Biological Rhythms. New York: PlenumPress,1982:499-522. 14. Broughton RG. Chronobiological aspects and models of sleep and napping. In Dinges DF, Broughton RG, eds. Sleep and alertness: Chronobiological Behavioral, and Medical Aspects of Napping, New York Raven Press, 1989:71-98. 15. Hartmann E. The 90 minute sleep-dream cycle. Arch Gen Psychiatry 1968; 18:280-286. 16. Brezinova V. Sleep cycle content and sleep cycle duration. Electroencephalogr Clin Neurophysiol 1974; 36:275-282. 17. Spiegel R. Analysis of the NREM-REM cycle. In: Weitzman ED, ed. Advances in sleep research, vol.5. Sleep and sleeplessness in advanced age.lancaster:mtp Press, 1981:153-168. 18. Kobayashi T,Tsujii Y, Endo S. Sleep cycle as a basic unit of sleep. Experimental Brain Res, Suppl.12, Heidelberg:Springer-Verlag Berlin, 1985:260-269. 19. Feinberg I, Fein G, Floyd TC. EEG patterns during and following extended sleep in young adults. Electroencephalogr Clin Neurophysiol 1980; 50: 467-476. 20. Dijk DJ, Cajochen C, Tobler I, Borbély AA. Sleep extension in humans: sleep stages, EEG power and body temperature. Sleep 1991; 14: 294-306. 21. Feinberg I. Changes in sleep cycle patterns with age. J of Psychiatry Res 1974; 10:283-306. 22. Borbély A, Baumann F, Brandeis D, Strauch I,Lehmann D. Sleep deprivation: effect on sleep stages and EEG power density in man. Electroencephalogr Clin Neurophysiol 1981; 51:483-493. 23. Borbély AA. A two process model of sleep regulation. Hum Neurobiol 1982; 1:195-204. 24. Feinberg I, March JD. Observations on delta homeostasis, the one stimulus model of NREM-REM alternation and the neurobiologic implications of experimental dream studies. Behavioural Brain Research 1995; 69: 97-108. 25. Schulz H, Dirlich G, Zulley J. Phase shift in the REM sleep rhythm. Pflugers Arch 1975; 358:203-212. 26. McPartland RJ, Kupfer DJ. Rapid eye movement sleep cycle, clock time and sleep onset. Electroencephalogr Clin Neurophysiol 1978; 45:178-185. 27. Carman GJ, Mealey L, Thompson ST, Thompson MA. Patterns in the distribution of REM sleep in normal human sleep. Sleep 1984; 7:347-355. 28. Zulley J. Distribution of REM sleep in entrained 24 hour and free running sleep-wake cycle. Sleep 1980; 2:377-389. 29. Lavie P. Ultrashort sleep-wake cycle: timing of REM sleep. evidence for sleep-dependent and sleep-independent components of the REM cycle. Sleep 1987; 10: 62-68. 30. Gillsberg M, Akerstedt A. The dynamics of the first sleep cycle. Sleep 1991; 14:147-154. 31. Feinberg I, Maloney T, March JD. Precise conservation of NREM period 1(NREMP 1) delta across naps and nocturnal sleep:implications for REM latency and NREM/REM alternation. Sleep 1992; 15:400-403. 32. Borbely A, Wirz Justice A.Sleep, sleep deprivation and depression. Hum Neurobiol 1982;1:205-210. 33. Feinberg I, Floyd TC, March JD. Effects of sleep loss on delta(03-3hz) EEG and eye movement density: new observation and hypotheses. Electroencephalogr Clin Neurophysiol 1987; 67:217-221. 34. Weitzman ED, Kripke DF, Pollak C, Dominguez J. Cyclic activity in sleep of macaca mulatta. Arch Neurol 1965; 12:463-467. 35. Kripke DF, Reite ML, Pegram GV, Stephens LM,Lewis OF. Nocturnal sleep in rhesus monkeys. Electroencephalogr Clin Neurophysiol 1968;24:582-586. 36. Ursin R. Sleep stage relations within the sleep cycles of the cat. Brain Res 1970; 20:91-9. 37. Tissot R. Essai methodologique d'analyse du cycle du sommeil Encephale 1966; 55:137-169. 38. Horne JA. Tissue restitution and sleep, with particular reference to human slow wave sleep. In S. Inoué and AA Borbely (eds.), Endogenous sleep substances and sleep regulation, VNU Science Press, Tokyo, 1984, SLEEP, Vol. 21, No. 3, 1998 275

pp. 25-40. 39. Achermann P, Dijk DJ, Brunner DP, Borbély AA. A model of human sleep homeostasis based on EEG slow-wave acivity: quantitative comparison of data and simulations. Brain Res Bull 1993; 31: 97-113. 40. Bunnell DE, Bevier WC, Horvath SM. Sleep interruption and exercise. Sleep 1984; 7:261-271. 41. Fukuda K, Miyasita A, Inugami M. Sleep onset REM periods observed after sleep interruption in normal short and normal long sleeping subjects. Electroencephalogr Clin Neurophysiol 1987; 67: 508-513. 42. Campbell SS. Evolution of sleep structure following brief intervals of wakefulness. Electroencephalogr Clin Neurophysiol 1987; 66: 175-184. 43. Campbell SS, Zulley J. Ultradian components of human sleep/wake patterns during disentrainment. Experimental Brain Res, Suppl.12, Heidelberg:Springer-Verlag Berlin, 1985:234-255. 44. Webb W, Agnew H. Sleep and waking in a time-free environment. Aerospace Medicine 1974; 45:617-622. 45. Nakagawa Y. Continous observation of EEG patterns at night and in daytime of normal subjects under restrained conditions. I; Quiescent state when lying down. Electroencephalogr Clin Neurophysiol 1980; 49:524-537. 46. Campbell SS. Duration and placement of sleep in a "disentrained" environment. Psychophysiology 1984; 21:106-113. 47. Coons S, Guilleminault C. Development of consolidated sleep and wakeful periods in relation to day/night cycle in infancy. Dev Med Child Neurol 1984; 26:169-176. 48. Wehr TA. A 'clock for all seasons' in the human brain. In Buijs RM, Kalsbeek A,Romihn JH, Pennartz CMA, Mirmiran M (eds.), Progress In Brain Research, Volume III, Elsevier, Amsterdam, 1996, pp. 319-340. 49. Harrison Y, Horne JA. Long term extension to sleep. Are we really chronically sleep deprived? Psychophysiology 1996; 33:22-30. 50. Borbely A, Achermann P. Concepts and models of sleep regulation: an overview. J Sleep Res 1992; 1:63-79. 51. Hume KI, Mills JN. Rhythms of REM and slow wave sleep in subjects living on abnormal time schedules. Waking Sleeping 1977; 1:291-296. 52. Czeisler CA, Zimmermann JC, Ronda JM, Moore-Ede MC, Weitzman ED. Timing of REM sleep is coupled to the circadian rhythm of body temperature. Sleep 1980; 2:329-346. 53. Carskadon MA, Dement WC. Distribution of REM sleep on a 90 minute sleep-wake schedule. Sleep 1980; 2:309-317. 54. Dijk DJ, Czeisler CA. Contribution of the circadian pacemaker and the sleep homeostat to sleep propensity, sleep structure, Electroencephalogralographic slow waves and sleep spindle activity in humans. The Journal of Neuroscience 1995; 15:3526-3538. 55. Garcia Borreguero D, Barbato G, Schwartz P, Wehr TA. Effects of suppressing the daily cortisol rhythm on the REM-sleep circadian rhythms. Neuropsychopharmacology 1994; 10:23S. SLEEP, Vol. 21, No. 3, 1998 276