Fragmenting Sleep Diminishes Its Recuperative Value

Transcription

1 Sleep 10(6): , Raven Press, Ltd., New York 1987 Association of Professional Sleep Societies Fragmenting Sleep Diminishes Its Recuperative Value Brian Levine, Timothy Roehrs, Edward Stepanski, Frank Zorick, and Thomas Roth Henry Ford Hospital, Sleep Disorders and Research Center, Detroit, Michigan, U.S.A. Summary: The recuperative effects of naps fragmented by different rates of electroencephalographic (EEG) arousal were evaluated. Forty healthy subjects with normal hearing and daytime sleep tendency (measured by the Multiple Sleep Latency Test at 10:00 a.m., 12:00 p.m., 2:00 p.m., and 4:00 p.m.) were randomly assigned to one of five conditions. Each was deprived of sleep for one night and then at 8:30 a.m. was given 100 min of natural sleep, sleep with arousals 1/5 min, 113 min, 111 min, or no sleep. After the recovery nap at 12:00 p.m., 2:00 p.m., 4:00 p.m., and 6:00 p.m., sleep latencies were again evaluated. Mean sleep latencies increased linearly as the rate of arousal during the recuperative nap decreased. Latency in the high-arousal condition was similar to no sleep and lower than natural sleep. The sleep latency of the lowarousal condition was similar to natural sleep and higher than no sleep, whereas latency in the medium arousal condition was intermediate to and differed from both natural sleep and no sleep. Although the natural sleep provided recuperation relative to no sleep or fragmented sleep, it did not restore daytime sleepiness to the screening level. Key Words: Excessive daytime sleepiness-sleep fragmentation-naps-eeg arousal. Perhaps the most common cause of excessive daytime sleepiness is the fragmentation of sleep. Sleep fragmentation refers to the punctuation of sleep with frequent, brief arousals characterized by increases in electroencephalographic (EEG) frequency or bursts of alpha activity, and occasionally, transient increases in skeletal muscle tone (1). These arousals last approximately 3-15 s, usually do not result in prolonged wakefulness, and sometimes may not even alter standard 30-s epoch sleep stage scoring. In some sleep disorders patients, the arousing condition or stimulus can be identified (i.e., apneas, leg movements, pain) (1-3). In other situations, the arousing stimulus has not been identified. For example, the sleep of healthy "normal" elderly is often fragmented (4), and out-of-phase sleep, such as occurs in shift work or jet lag, also is fragmented (5). Although it is recognized that the fragmentation of sleep results in excessive daytime sleepiness, the phenomenon has been the subject of relatively few systematic studies. Address correspondence and reprint requests to Dr. Timothy Roehrs at Henry Ford Hospital, 2921 west Grand Boulevard, Detroit, MI 48202, U.S.A. 590

2 SLEEPINESS AND FRAGMENTATION 591 1' Several correlational studies have shown a relation between sleep fragmentation and daytime symptoms. Fragmentation of sleep, as indexed by number of brief EEG arousals, number of shifts from other sleep stages to stage 1 sleep or wake, and the percentage of stage 1 sleep have been found to correlate with excessive daytime sleepiness (3,6). Patients with sleep apnea syndrome who are successfully treated by surgery (i.e., apnea is reduced) show a decrease in the number of arousals from sleep and also in excessive daytime sleepiness, whereas those not benefiting from the surgery (i.e., apneas remained) do not show a decrease in arousals and a concomitant decrease in excessive daytime sleepiness (7). A recent study of the recuperative effects of naps after sleep deprivation found naps of min increased alertness, but additional sleep (120 min) provided no further refreshment (8). The additional 60 min of sleep on the 120 min nap was fragmented with arousals, whereas the first 60 min were free of arousals, as was the sleep of the 15-, 30-, and 6O-min naps. Sleep fragmentation has been experimentally studied by inducing arousals in normals with external stimuli. Several studies have employed an auditory stimulus to awaken subjects at various intervals during the night (9-11). Decrements in performance and a single sleep latency test were related to the "periodicity of disturbance" and not to sleep staging variables. However, requiring subjects to respond behaviorally to the auditory stimulus results in awakening that is somewhat dissimilar to the experience of patients with excessive daytime sleepiness who do not awaken completely. In another study, tones were presented to subjects during the night at 5.5-min intervals, and subsequent increases in daytime sleepiness were observed, without increased wakefulness during the sleep period (12). This was accomplished by terminating the tones upon arousal (a speeding of the EEG or a burst of alpha activity of at least 3 s in duration), not behavioral wakefulness. Also, sleepiness was measured repeatedly throughout the day with the Multiple Sleep Latency Test (MSLT), which has been shown to be a reliable measure of daytime sleepiness. It is systematically related to the amount of prior sleep (in deprivation and sleep restriction studies) and the effects of sedating and alerting drugs (13-16). Yet, there were some methodological problems, including a single rate of arousals, a small "n", and the use of a repeatedmeasures design, which may have contributed to a rapid habituation to the arousal methodology. The studies above do provide preliminary experimental evidence indicating that sleep fragmentation produces daytime sleepiness, but they raise further questions as to how arousal rate relates to daytime sleepiness. It is presumed that sleep serves some restorative function, which in turn makes it possible to maintain full alertness during subsequent periods of wake. Furthermore, arousals during sleep disrupt its continuity and somehow impair the restorative process. However, it is difficult to compare arousal rates among the different studies, each with different methodologies, to determine how arousal rate affects the restorative process. No single study has attempted a parametric analysis of arousal rate and daytime sleepiness. This study further evaluated the effects on subsequent daytime sleepiness of fragmenting sleep by brief arousals. It used the recuperative nap methodology in normals to assess the effects of fragmenting sleep at different rates of arousal. METHODS Subjects Twenty men and 20 women between the ages of 18 and 35 years participated. All subjects were healthy, nongravid, reported no excessive alcohol consumption or the Sleep, Vol. 10, No.6, 1987

3 592 B. LEVINE ET AL. use of illicit drugs, and were not using medications that affect sleep. They reported no sleep problems, no daytime sleepiness, and did not nap. Their hearing was normai, based on standard audiometric testing, and their sleep latency was normal, based on a screening MSL T. Subjects signed informed consents to participate and were paid for their participation. Design The 40 subjects were randomly assigned to one of five conditions. Subjects in each condition had one night of sleep deprivation followed by a 100-min recovery nap in four of the conditions. The first condition had no sleep, the second a nap with arousals programmed once per minute, the third with arousals once per 3 min, the fourth with arousals once per 5 min, and the fifth condition had a nap with no experimentally induced arousals. After the recovery nap, sleep latency tests were performed at 12:00 p.m., 2:00 p.m., 4:00 p.m., and 6:00 p.m. to evaluate daytime sleepiness. Procedure On the first visit to the laboratory, subjects were fully instructed regarding study procedures (including possible disruption of sleep during a nap), signed the Informed Consent sheet, and completed the Cornell Medical Index. An air-conduction threshold hearing test was administered by a certified audiologist on equipment calibrated to ANSI standards (1969) in a sound-attenuated room. Thresholds were obtained at discrete frequencies of 250, 500, 1,000, 2,000, 4,000, and 8,000 Hz. On another day, the screening MSLT was administered. Each subject slept at the laboratory the night prior to their MSLT, spending 8 h in bed, beginning at their usual bedtime. Before going to bed, they filled out a brief questionnaire that asked if they had any naps, alcohol, or caffeine during the day. During the screening night, subjects were monitored by a wrist-worn activity monitor or standard polysomnography, depending on availability of polygraphs. Subjects were awakened 8 h after lights out. They were allowed to breakfast either immediately after awakening or after the 10:00 a.m. latency test. If desired, lunch was taken after the 12:00 p.m. latency test. Between 9:00 and 10:00 a.m., electrodes were attached for a standard five-channel sleep recording [central and occipital EEG, horizontal electrooculogram (EOG), and submental electromyogram (EMG)] using the Rechtschaffen and Kales montage (17). At 10:00 a.m., 12:00 p.m., 2:00 p.m., and 4:00 p.m., subjects went to bed in a dark room for the latency tests. They were told to close their eyes, relax, and to sleep. Latency tests were terminated at first signs of sleep (at least two 30-s epochs) or 20 min of wakefulness. The subjects were observed between the scheduled tests and were not allowed to lie down or to fall asleep at other than scheduled times. At no point in the study were the subjects told that their sleepiness was being measured as a criterion for inclusion in the study. All subjects with an average latency of 8 min or greater were asked to come back for the second night. On the second night, subjects reported to the sleep laboratory at 10:00 p.m. During the night, they were allowed to do homework, play games, or watch television. A monitor was appointed to keep them awake and in the sleep laboratory. In the morning, they were prepared for a standard sleep recording as described above. Subjects were allowed to breakfast anytime before 8:00 a.m. At 8: 15 a.m., Realistic in-ear headphones (model no ) were placed in the ears of subjects in the four sleeping conditions and reinforced with tape. These small, unobtrusive headphones were of minimal discomfort to the napping subjects. At 8:30 a.m., subjects were allowed to Sleep, Vol. 10, No.6, 1987

4 SLEEPINESS AND FRAGMENTATION 593 sleep. Upon the first signs of stage 2 sleep, 20 epochs (10 min) of uninterrupted sleep were counted before initiating the tones. This was done to ensure that subjects in the different arousal conditions equally achieved consolidated sleep before initiating the fragmentation. Tones were generated by a tone generator interfaced to an Apple computer. The generator was calibrated to emit a 90-dB tone at 625 Hz. The 90-dB level was determined from previous data that indicated that arousal thresholds rose to 90 db after repeated presentation of tones over two nights (12). The desired average rate of the tones (1/1 min for the high-frequency condition, 113 min for the medium-frequency condition, 115 min for the low-frequency condition) was programmed before the session began. The program presented tones at random intervals not more than ± 20% of the selected interval (e.g., the low-frequency condition received tones every 4-6 min, with an average of 5 min). The headphones worn by subjects in the natural sleep condition were not connected to the tone generator. Each presentation of the tones consisted of one I-s tone every 10 s for three trials, and every 2 s thereafter. If the subject was awake by Rechtschaffen and Kales criterion (17) for 30 s before a tone presentation, the presentation was aborted and postponed until 30 s of sleep had elapsed. Termination of tones was done by the experimenter upon signs of arousal according to EEG criteria. The necessary EEG criteria were an increase in EEG frequency or a burst of alpha activity at least 3 s in duration (12). In extreme cases, it was necessary to present a novel stimulus (1-10 rapid tones) over the room intercom in order to produce an arousal. This occurred most often in the high-frequency condition. As the nap progressed, the number of epochs of wake were tallied. For each epoch of wake, subjects were allowed to sleep an extra epoch in order to equalize total sleep time among all four napping groups. Thus, bed time on the nap ranged among conditions from 105 to 115 min on average, depending on latency to stage 2 and total wake time. After the 8:30 a.m. nap, subjects were awakened, and the headphones were removed from their ears. Latency tests were administered at 12:00 p.m., 2:00 p.m., 4:00 p.m., and 6:00 p.m. Lunch, if desired, was taken after the 12:00 p.m. latency test. The protocol for the latency tests was identical to that of the screening MSLT, with the exception of the later test times (12:00-6:00 p.m.), and subjects were told that they would be taking a brief symbol copying test at 6:45 p.m., about 30 min after the last latency test. This was done in order to eliminate any end-of-experiment restlessness that may confound the last latency test. All sleep records were scored by the experimenter according to standards established by Rechtschaffen and Kales (17). Sleep latency on the latency tests was scored at the first of two consecutive epochs of stage 1 sleep or one epoch of another sleep stage. Arousals were defined as a speeding of the EEG or the presence of alpha activity at least 3 s in duration and possibly accompanied by a discrete elevation of the submental EMG. Linear regression analyses were used to determine the nature of the relation of sleep latency to napping conditions. Between-groups and mixed-design analyses of variance with Greenhouse-Geisser corrections and post hoc Duncan multiple range tests were performed, comparing the sleep latencies (mean sleep latency and latency on each individual test) among the five conditions and comparing the sleep parameters of the four napping groups. Pearson product correlations were used to assess the relation between the various sleep parameters and mean sleep latency and between the number of arousals and mean sleep latency. Sleep. Vol. 10, No.6, 1987

5 594 B. LEVINE ET AL. RESULTS Sleep Deprivation Effects There were no significant differences among the five conditions in mean sleep latency (of each subject's four tests) on the screening MSLT. These data are presented in Table 1. Mean latency varied among conditions from 12.6 to 15.1 min, and the typical circadian pattern of sleep latency with a nadir in latency over the midday was observed (F = 4.78; df 3,105; p < 0.005; 10:00 a.m. versus 2:00 p.m.). On the morning (8:30 a.m.) nap following the one night sleep deprivation, sleep latency did not differ among conditions (see Table 1). The nap latency was between 1.88 and 1.31 min among conditions, with standard deviations of approximately 1 min. These sleep deprivation effects persisted throughout the day, as can be seen in the no-sleep condition. Figure 1 presents the latency for each condition on each test over the day. Mean latency in the no-sleep condition was 2.2 min, and on each test, the latencies remained between 2 and 3 min. The latencies in the no-sleep condition were significantly reduced compared to the screening day at comparable (12:00, 2:00,4:00 p.m.) times (F = 135; df 1,7; p < 0.00). Arousals on the Recuperative Nap The number of experimental and natural arousals in each condition is presented in Table 2. With the exception of the high-frequency condition, experimental arousal rate was close to the originally intended rates. Subjects in the high-frequency condition had experimental arousals (compared to 90 intended), the medium-frequency condition had (compared to 30), and the low-frequency condition had (compared to 18). Although subjects in the natural sleep condition did not have any experimentally induced arousals, their sleep was somewhat fragmented with natural arousals (mean 8.13). Also, the sleep in the other conditions showed natural arousals; the number of natural arousals did not differ among the conditions (see Table 2). With the additional TABLE 1. Sleep latencies (min) on screening MSLT and at 8:30 a.m. postdeprivation Conditions High Medium Low Natural No sleep 1/1 min 1/3 min 1/5 min sleep Screening MSLT Mean (2.42) (2.08) (2.59) (2.98) (2.95) 10:00 a.m (4.99) (3.95) (2.94) (5.57) (3.46) 12:00 p.m (5.82) (4.59) (5.10) 0.58) (3.48) 2:00 p.m (4.52) (5.29) (4.76) (3.76) (4.61) 4:00 p.m (6.12) (5.35) (4.57) (3.25) (4.68) Latency at 8:30 a.m. postdeprivation \ (1.22) (\,22) (1.00) (1.28) Data are means with standard deviations in parentheses. 1/5 min: a nap with an arousal programmed every 5 min; 1/3 min: an arousal every 3 min; 111 min: an arousal every min. Sleep, Vol. /0, No.6, 1987

6 SLEEPINESS AND FRAGMENTATION 595 ;---. VJ C I 10 8 >- 6 U z w ~ 4 a. W W -l (f) 2 a Conditions 0-0 Not Sip e-e 1/5 min 6-61/3 min a / &-o~..-.1/1 min 0-0 No Sip ~:2~~~ HOUR FIG. 1. Sleep latency on each latency test for each recuperative nap condition. Nat Sip, natural sleep; 115 min, a nap with an arousal programmed every 5 min; 1/3 min, an arousal every 3 min; 1/1 min, an arousal every min; and No Sip, the no-sleep condition. natural arousals and the total sleep times, as shown in Table 3, the actual arousal rates were 1 per 1.4, 2.5, and 4.2 min in the high, medium, and low frequency conditions, respectively. Interestingly, subjective reports of awakenings over the approximate 100-min nap were very low. Subjects reported awakening on average 4.57 times for the high-frequency condition, 6.71 times for the medium-frequency condition, 1.14 times for the low-frequency condition, and 0 times for the natural sleep condition (see Fig. 2). The number of subjective reports of awakening did not differ among the four conditions. Recuperative Effects of Different Nap Conditions Mean sleep latency (of each subject's four tests) was an increasing function of the reduced frequency of arousal rate. Figure 3 presents the mean sleep latency of the four latency tests for each of the nap conditions. A regression analysis of the mean sleep latency for the five conditions revealed that the linear polynomial was significant (F = 8.11; df 3; p < 0.01), whereas the cubic and quadratic polynomials were not significant. Significant linear polynomials were also found for latency on the 2:00 p.m. test (F = 5.69; df 3; p < 0.01), the 4:00 p.m. test (F = 5; df 3; p < 0.01), and the 6:00 p.m. test (F 5.20; df 3; p < 0.01). For the 12:00 p.m. latency test, the medium-frequency condi- TABLE 2. Natural EEG arousals and experimental EEG arousals Conditions High Medium Low Natural Arousal type No sleep III min 113 min 115 min sleep Natural (5.35) (7.54) (7.39) (3.68) Experimental (7.13) (1.91) (2.77) (0.00) Total arousals (5.63) (7.27) (7.09) (3.68) Data are means with standard deviations in parentheses. Sleep, Vol. 10, No.6, 1987

7 596 B. LEVINE ET AL. TABLE 3. Sleep parameters on the recuperative nap Conditions High Medium Low Natural Parameter No sleep 111 min 1/3 min 1/5 min sleep TIB (11.23) (6.95) (4.90) (3.82) TST (7.58) (4.64) (3.69) (3.37) Percent stage I (19.37) (16.78) (5.00) (4.96) Percent stage ) (13.90) (11.06) (14.90) Percent stage 3/ (24.85) (9.65) (16.69) (9.14) Percent stage REM (6.14) (8.36) (8.45) (7.88) Sleep efficiency (5.31) (4.58) (2.98) (1.47) Data are means with standard deviations in parentheses. tion had a higher latency than the low-frequency condition. None of these latency tests had significant cubic or quadratic polynomials. Mean sleep latency (of each subject's four tests) differed significantly among the five conditions (F = 5.92; df 4,35; p < 0.01). The no-sleep condition had a shorter mean latency than the medium-frequency, low-frequency, and natural sleep conditions, but was not significantly shorter than the high-frequency condition. The natural sleep condition had a longer mean latency than the no-sleep and the high- and medium-frequency conditions, but was not significantly longer than the low-frequency condition. The high-, medium-, and low-frequency conditions did not significantly differ. An analysis of the four separate latency tests (mixed design ANOYA with condition as the between-group factor and time of test the within-group factor) also was conducted. The main effects of condition (F = 5.73; df 4,35; p < 0.01) and time (F = 3.12; df 4,106; p < 0.04) were significant. As expected from the omnibus analysis, separate (f) -l «(f) ::> 0 c:: «ll. 0 '* lzz2i Total EEG Arousals _ Reported Arousals No Sip 1/1 min 1/3 min 1/5 min Nat Sip CONDITIONS FIG. 2. Total number of EEG arousals and reported awakenings in each recuperative nap condition. Abbreviations as in Fig. 1. Sleep. Vol. 10, No.6, 1987

8 SLEEPINESS AND FRAGMENTATION ,,... c 5 8 b ~ 6 ~ ~ 4 w...j Ul z 2 L5 :::; No Sip 1/1 min 1/3 min 1/5 min Nat Sip CONDITIONS FIG. 3. Mean sleep latency of the four latency tests following each of five recuperative nap conditions. Abbreviations as in Fig. I. analyses of each sleep latency test each yielded a significant conditions effect. However, the ordinal relation among conditions on the separate tests was somewhat inconsistent (see Fig. 1). On the 12:00 p.m. test, the only significant differences were between the natural sleep and no-sleep conditions and the medium-frequency and the no-sleep conditions. On the 4:00 p.m. test, unlike the other tests, the high-frequency condition differed from the no-sleep condition. The relation between fragmentation on the nap and subsequent sleepiness was also assessed by correlating number of arousals and sleep latency. There was a significant correlation (r = , p < 0.05) between number of arousals and mean sleep latency. Sleep Stages and Recuperation As might be expected to result from the experimental fragmentation of sleep, sleep efficiency significantly differed among conditions (F = 6.59; df 3,28; p < 0.0l). The medium-frequency condition had the lowest sleep efficiency (90.08%), significantly lower than the low-frequency condition (96.23%) and the natural sleep condition (97.74%). The high-frequency condition had a lower sleep efficiency (92.33%) than the natural sleep condition (see Table 3). There was no significant correlation between sleep efficiency and subsequent sleep latencies. The most notable differences in sleep staging were in stage 1 sleep (F = 18.70; df 3,28; p < 0.01). The high-frequency condition had a significantly higher percentage of stage 1 sleep than any other condition (53.15%). The medium-frequency condition had a higher percentage of stage 1 sleep (36.34%) than the low-frequency condition (19.35%) or the natural sleep condition (6.81%). The natural sleep condition had 47.14% stage 3-4 sleep, which is significantly more than any other condition (F = 8.56; df 3,28; p < 0.01). Stage 2 and rapid eye movement (REM) did not differ among conditions. The only sleep stage to correlate significantly with sleep latency score was stage 1 (r = -0.41, p < 0.05). As for other sleep parameters, time in bed (TIB) and total sleep time (TST) did not differ among the four napping conditions (see Table 3). DISCUSSION This study makes several important additions to our present understanding about the relation between sleep fragmentation and daytime sleepiness. First of all, using a dif- Sleep, Vol. 10, No.6, 1987

9 598 B. LEVINE ET AL. ferent methodology (the recuperative napping methodology), this study confirms the correlational studies and the several experimental studies that indicate that fragmentation of sleep by EEG arousals without sleep loss per se is related, to daytime sleepiness. Secondly, the relation between rate of arousal and subsequent sleepiness was linear. The linearity implies that within the range of arousal rate tested in this study (about min), there is no single arousal rate about which sleep is either restorative or nonrestorative, but rather, as the arousal rate increases, the restorative capacity of sleep is diminished. Several points can be made to support the internal validity of the results of this study. Sleep latency at 8:30 a.m., as well as TST, TIB, and screening latency, did not differ among the conditions. Therefore, the effects of differing levels of initial sleepiness, extra sleep, extra TlB, or differential baseline sleepiness did not confound the effects of varying rates of arousals. Furthermore, to demonstrate the effects of varying rates of arousals, it was necessary to show that the 100-min nap (without tones) was sufficient to achieve measurable alerting effects versus no sleep. This was demonstrated, partially replicating a previous study in which a 60- and l20-min nap significantly raised latencies above no sleep (8). Even though subjects who received tones had arousals, they only remembered a small fraction of them. This is probably related to the brief duration of the arousals, which, due to the excessive sleepiness caused by sleep deprivation, was rarely long enough to be classified as wakefulness. Further, behavioral awakening was not a criteria of arousal in this study either, only EEG signs of arousal. Consequently, these results are not confounded by reductions in sleep time among conditions. In fact, additional sleep was allowed if necessary to equalize sleep time among conditions. It was anticipated that sleep staging would be affected by sleep fragmentation. It is not surprising that stage 1 was significantly different among conditions and that stage 1 significantly correlated with sleep latency following the nap. Other studies have found percentage of stage I sleep to be an indicator of the severity of sleep fragmentation and subsequent sleepiness. Stage 3/4 was not predictiw,,1 daytime sleepiness, and the only difference was the natural sleep condition, which differed from all others. Sleep efficiency, like sleep staging, was expected to differ as a function of sleep fragmentation. However, the sleep efficiency differences among conditions were not as extensive nor as systematic as sleep staging differences (e.g., the medium-frequency condition had the lowest sleep efficiency and differed from the others), yet, all conditions had sleep efficiencies greater than 90%. Therefore, sleep efficiency never really reached highly impaired levels, and furthermore, it did not significantly correlate with sleepiness. There was enough fragmentation in the high-frequency condition to increase sleepiness above that of undisturbed sleepers, unlike the low-frequency condition, which was not sleepier than the natural sleep condition. This suggests that the point at which sleep becomes so fragmented as to have no alerting effect lies somewhere between 30 and 65 arousals per 90 min. A correlate to this observation is that units of sleep at least as short as 3 min in length on average are still long enough to have an alerting effect, even though they are not as alerting as undisturbed sleep. Units of sleep 5 min in length were as alerting as undisturbed sleep. These findings regarding the restorative effects of different rates of fragmentation must be interpreted with some caution. In the first place, these data were collected on an approximate 100-min nap, and they may not apply to a full8-h sleep. The low rate of arousal, which showed no impairment of restoration in this study, may produce cumu- Sleep, Vol. 10, No.6, 1987

10 SLEEPINESS AND FRAGMENTATION 599 lative effects over an 8-h sleep period relative to natural sleep. In fact, two full-night studies showed effects with rates of 115 min and 1110 min (9,12). Secondly, we were unable to produce true "dose effects" in the sense of differences between the three conditions with experimentally induced arousals. The failure to produce "dose" differences probably occurred because there were not enough arousals in the high-frequency condition. The subjects in the high-frequency condition were more difficult to arouse, and therefore, fewer arousals than intended could be induced in the 90-min period. In addition, higher than desired within-condition variability of arousal number may also have produced overlap among conditions. In other words, the arousal rates (especially in the low- and medium-frequency conditions) did not differ enough. In summary, these data clearly do indicate that the rate of arousal during sleep is systematically related to the restorative effects of that sleep. That relation is linear, with increasing arousal rate diminishing sleep's restorative capacity. It remains to be determined as to how these relations will generalize to a full 8-h night's sleep. Acknowledgment: This research was supported by NIH Grant ROI NS , awarded to Dr. Roth, and by NIH Grant ROI NS , awarded to Dr. Roehrs. REFERENCES 1. Roth T, Hartse KM, Zorick F, Conway W. Multiple naps and the evaluation of daytime sleepiness in patients with upper airway sleep apnea. Sleep 1980;3: Rosenthal L, Roehrs T, Sicklesteel J, Zorick F, Wittig R, Roth T. Periodic leg movements during sleep. Sleep 1984;7: Miles L, Dement WC. Sleep pathologies. Sleep 1980;3: Carskadon MA, Brown E, Dement WC. Sleep fragmentation in the elderly: relationship to daytime sleep tendency. Neurobiol Aging 1982;3: Wegman H, Gundel A, Nauman M, Same! A, Schwartz E, Vejvoda M. Sleep, sleepiness, and circadian rhythmicity in aircrews operating on transatlantic routes. Aviation Space Environ Med 1986;57:B53- B Stepanski E, Lamphere J, Badia P, Zorick F, Roth T. Sleep fragmentation and daytime sleepiness. Sleep 1984;7: Zorick F, Roehrs T, Conway W, Fujita S, Wittig W, Roth T. Effects ofuvulopalatopharyngoplasty on the daytime sleepiness associated with sleep apnea syndrome. Bull Eur Physiopathol Respir 1983;19: Lumley M, Roehrs T, Zorick F, Lamphere J, Wittig R, Roth T. Alerting effects of naps in normal sleep deprived subjects. Psychophysiology 1986;23: Bonnet MH. The effect of sleep disruption on performance, sleep, and mood. Sleep 1985;8: Bonnet MH. Performance and sleepiness as a function of frequency and placement of sleep disruption. Psychophysiology 1986;23: Bonnet MH. Cumulative effects of sleep restriction on daytime sleepiness. Physiol Behav 1986;37: Stepanski E, Lamphere J, Roehrs T, Zorick F, Roth T. Experimental sleep fragmentation and sleepiness in normal subjects. Int J Neurosci 1987;33: Carskadon MA, Harvey KM, Dement WC. Sleep loss in young adolescents. Sleep 1981;4: Carskadon MA, Dement WC. The multiple sleep latency test: what does it measure? Sleep 1982;5:S Seidel W, Roth T, Roehrs T, Zorick F, Dement WC. Treatment of a 12-hr shift of sleep schedule with benzodiazepines. Science 1984;224: 1, Roth T, Roehrs T, Zorick F. Sleepiness: its measurement and determinants. Sleep 1982;5:S Rechtschaffen A, Kales A (eds). A manual of standardized terminology, techniques and scoring system for sleep stages of human adults. Los Angeles: Brain Information Service/Brain Research Institute, UCLA,1968. Sleep, Vol. /0, No.6, 1987