- Email: [email protected]
Characteristics of rapid eye movement production during rem sleep in man: Organization and rhythmicity
Electroencephalography and clinical Neurophysiology , 1983, 55:151-155 Elsevier Scientific Publishers Ireland, Ltd.
151
CHARACTERISTICS OF RAPID EYE MOVEMENT PRODUCTION DURING REM SLEEP IN MAN: ORGANIZATION AND RHYTHMICITY L.D. WEBER, A. MUZET, J.P. SCHIEBER and J.P. LIENHARD Centre d'Etudes Bioclimatiques du C.N.R.S., 21 rue Becquerel, 67087 Strasbourg C E D E X (France)
(Accepted for publication: October 1, 1982)
Recently, researchers have become interested in certain characteristics of rapid eye movement (REM) production during REM sleep in man. Several studies have been published concerning REM organization within a REM sleep phase. For example, Aserinsky (1971) found that in REM sleep phases equal to or greater than 20 min, ocular motor activity peaked 5-10 rain after the onset of REM sleep and then decreased 10 min later. Salzarulo (1972) reported that REM density variations showed a peak value, the location of which within the phase differed according to the length of the phase. He further stated that this density peak was reached more quickly in longer phases of REM sleep and that after 21 min of REM sleep, REM density always decreased progressively and regularly to reach a value near zero. Others have studied the rhythmic or periodic aspect of REM production. Krynicki (1975) found a rhythmic production of REMs whose period ranged from 10 to 30 min. Lavie (1979), however, found that this periodicity was limited to between 10 and 20 min. The purpose of the present study was to verify an organized and rhythmic production of REMs by means of a mathematical technique which would allow us to test individual spectral peaks that might be present in series of REMs.
Method
Six young adults, of both sexes, between 19 and 24 years of age, spent 20 consecutive nights in the sleep laboratory. During this time these subjects took part in a study concerning the effects of noise
on the sleep of different age groups. The first 3 nights were reference nights, the following 15 nights were disturbed by traffic noises and the last 2 nights were recovery nights. Lights out was at 23.00 h and subjects remained in bed until 07.00 h the following morning. During the reference nights (N1, N2), the two first and last disturbed nights (N3, N4, N16, N17) and the recovery nights (NI8, N I9), the following uninterrupted electrophysiological records were obtained: EEG (derivations F3-A2 and C3-A2); 2 electro-oculograms (EOGs; the electrodes were placed slightly above and below the outer canthus); EMG of either the mentalis or the masseter; ECG; finger pulse amplitude; pulse wave velocity; body mobility, measured by a bed actograph and a radar detection system. During the remaining noise-disturbed nights (N5N15), only cardiovascular records were made. The noise emitted during the disturbed nights of this study consisted of 3 traffic noises whose intensities were 45, 55 or 65 dBA and whose durations were 4, 8 and 12 sec respectively. These noises were continuously diffused during the night according to a semi-random temporal distribution at a rate of 90 noises/h. Other nocturnal ambient conditions were maintained constant for all experimental nights (air temperature of 20°C _+ I°C; a relative humidity of 60% _+ 5%; a background noise due to air conditioning of 35 dBA). All records were visually scored for sleep stages in 30 sec epochs, according to the criteria given by Rechtschaffen and Kales (1968). REMs were detected visually and counted by periods of 30 sec. Only REMs that were visible in both EOGs and with an amplitude of at least 5 mm (60 ~tV) in one of the two channels were taken into consideration
0013-4649/83/0000-0000/$03.00 © 1983 Elsevier Scientific Publishers Ireland. Ltd.
152
L.D. WEBER ET AL.
for this study. REM sleep phases which were less than 20 min in length, interrupted by more than 5 min of N R E M sleep or interrupted by experimenter intervention were not analysed. In all, a total of 114 REM sleep phases which occurred during the 3 reference nights, disturbed nights N3, N4, N16 and N17, and the 2 recovery nights (nights for which we had EOG records) qualified for statistical analysis.
Statistical analysis Estimation and subtraction of slow trends of REMs The method which was used consisted of adjusting a least squared polynome to our series of REMs. This was achieved by calculating the coefficients of 8~ for i = 0, 1 ... K of a polynome of K degree (Pi = 80 + ~ l X i -at-t~2 x 2 " ' " + ~ K X K ) • The polynomial values which were calculated for each i were then subtracted point by point from the observed series of REMs (formation of the differences: ~ ( i = X i - P i ) . The curve which was traced from the Pi polynomial values indicates the evolving tendency or slow trend of the observed series. The )(i values were then used to find possible periodic fluctuations in comparison with the slow trend. Periodicity or quasi-periodicity detection by a power spectral method The method of Fourier transformation of the autocorrelation function and its associated errors of estimation has been described by Blackman and Tukey (1958). The power spectrum which is obtained in this way must be corrected by adequate smoothing of the calculated spectral values (p.S.xi p. S.x, : p.s. = power spectrum; - indicates smoothing). Under these conditions p.'s. can be considered as an acceptable estimate of the real power spectrum of the series X i, or as a distribution of the total variance of the observed series in independent fractions, in a certain number of frequency bands. This property (Parseval theorem) is important because it allows for the possibility of testing for the existence of periodicity (significant variance) in one or more given spectral bands. ~
Thus, the independent fractions of the variance calculated for the different spectral bands can be compared among themselves either individually or cumulatively by comparing and weighting them with the associated degrees of freedom (Snedecor F). With this method, the longest rhythmic periods that can be detected are limited to N x 2, while the shortest periods are equal to 2 x the sampling interval (Shannon rule). This method was used to test the estimated values of variance in certain frequency bands where the variance appeared to be significantly higher than the 'residual' variance located in the remaining frequency bands of the spectrum. An example of this analysis can be found in Table I.
TABLE I Example of a spectrum of variance showing 2 zones (boxes) of cumulated spectral values that were visually detected and then tested for their statistical significance (Snedecor F test). The first spectral value was always eliminated, due to the accumulation of a large part of the variance in this part of the spectrum which corresponds to low frequencies. Frequencies are expressed in c/min. Frequency
Spectral value
0.0556 0.1111 0.1667 0.2222 0.2778 0.3333 0.3889 0.4444 0.5000 0.5556 0.6111 0.6667 0.7222 0.7778 0.8333 0.8889 0.9444 1.0000
4.0772
** P < 0.01.
1.1765 [~.392~ 12.3885 I 1.2401 0.8177 /0.7216 0.4453 0,5861 0.8954 0.8791 0.6932 0.5860 0.8463 1.0538 0.9901
F
df
1.39
16/127
2.33
16/127
S
**
R E M s D U R I N G REM SLEEP IN M A N
153
Results
60
Of the 114 REM sleep phases that were analysed: 11 were 1st REM sleep phases of the night, 34 were 2nd, 33 were 3rd, 29 were 4th and 7 were 5th.
--
B
50
40
~~~~
30
Slow trend analysis Results showed that there were 3 types of predominant REM production trends in our series of REMs. These evolving trends, along with their frequency of occurrence and the mean duration of the REM sleep phase during which they occurred, can be found in Table II. The occurrence of a given slow trend was not related to subject, experimental night or time of night. In addition, the slow trend form was not related to the length of the REM sleep phase, with the exception of those eye movement series that were shown to have no significant trend. In this case REM sleep phases were short in duration. However, in 20 REM phases that were shorter than 25 min, only 6 had no significant slow trends while the remaining 14 did. Detection of periodicity of rapid eye movement production The spectral analysis of REMs was made after having subtracted the slow trends of REMs. Afterwards, during both non-disturbed and disturbed nights, all dominant spectral values were visually detected and noted. Finally, these spectral values were tested by means of a Snedecor F test in order to expose those spectral values which were statistically significant. Results showed that during the
T A B L E II Types of slow evolving trends present in rapid eye movement series equal to or longer than 20 rain. Slow trend type
Frequency of occurrence
Per cent
Mean duration (min)
Quadratic Quasi-sinusoidal Linear Other No trend
52 28 20 8 6
46 25 17 7 5
40.9 37.3 42.6 39.2 21.6
visuaLLy detected rhythmic peaks N:269 statisticaLLy sigmficant rhythmic periods N:111
20 10 0
5
10
15
20 m;n
Length of rhythmic periods
Fig. 1. Histograms of spectral estimations (in rain). A distinction has been made between spectral peaks that were visually detected and spectral peaks which were statistically significant (Snedecor F test). Non-disturbed and disturbed nights have been cumulated.
non-disturbed nights, only 40% of the spectral values that were visually detected were in fact statistically significant. This proportion was 41% for the disturbed nights. During the non-disturbed nights, REMs were modulated by statistically significant ( P < 0.01) rhythmic periods which were inferior or equal to 3 rain in 14% of the cases, ranging between 3 and 6 rain in 33% of the cases, ranging from 6 to 9 min in 18% of the cases and superior to 9 min in 35% of the cases. During the disturbed nights these proportions were: 18%, 35%, 8% and 39% respectively. Because of the large similarity between the distribution histograms of the rhythmic peaks during the non-disturbed nights and disturbed nights, we cumulated our observations in order to increase the quantity of our data. The corresponding histogram can be found in Fig. 1.
Discussion
Our results are in agreement with some of the conclusions reached by Aserinsky (1971), Krynicki (1975) and Lavie (1979), in that they support the
154
notion of an organized as opposed to a random distribution of REMs during REM sleep. Slow trend detection in our series revealed several predominant trend types. Like others (Salzarulo 1972; Krynicki 1975) we found quadratic and linear type trends, in addition to quasi-sinusoidal type trends. The occurrence of a given trend was not dependent on subject, night, time of night or length of the REM sleep phase. We found, as did Krynicki (1975), that at least one of the quadratic trends presented a concave-up form rather than the more common concave-down form, and that at least two of the linear trends were increasing rather than decreasing. The explanation of the presence of these different production trends in our series of REMs is for the moment speculative. Nonetheless, in the light of our findings and a certain number of those in the literature, we are led to the following hypothesis. It is possible that these different slow trends may be in actuality the expression of parts of the same, continuous oscillating process controlling ocular motor activity, this process being present throughout the night. If we admit such an underlying process, we should then propose that REM sleep phases would occur during different parts of this process, thus accounting for the presence of different slow trend types. This hypothesis would suggest that while the appearance of REMs during REM sleep would be controlled by a REM sleepdependent mechanism, the characteristics of the production of REMs (trend type and rhythmic period) would be controlled by a mechanism of another nature, being independent of REM sleep. Our hypothesis supports the findings of Zimmerman et al. (1980), who found that REM density was dissociated from REM sleep timing under free-running sleep conditions. Their results showed that while there existed differences in REM sleep accumulation and duration in entrained and free-running sleep conditions, there were no such differences in REM production patterns. Further, our hypothesis supports those made by Aserinsky (1969, 1973) in which he related increases in REM density during the night to the amount of prior accumulated sleep. Before discussing the results of the detection of rhythmic or periodic production of REMs during
L.D. WEBER ET AL.
REM sleep, we would like to specify what we think are two important points concerning the method used in this analysis. Firstly, certain authors chose not to subtract the slow trend from REM series before the spectral analysis of these series. The inconvenience of this method is the accumulation of a large part of the variance in that part of the spectrum which corresponds to low frequencies. The origin of this 'energy' (variance) is not necessarily of a harmonic nature. In addition, the presence of this energy, often relatively large in relationship to the rest of the spectrum, makes the detection of rhythms by analysis of variance less sensitive. For these reasons we found it preferable to eliminate the slow trend. Secondly, it is possible that a series of REMs may contain one, two or more statistically significant rhythmic periods, or no significant rhythmic period. Thus, we found it necessary to test all visually detected spectral peaks for their statistical significance. As can be seen from our results, approximately 60% of the spectral peaks which were visually detected in our series of REMs were not statistically significant. These two points concerning the analytical method which we used in detecting the rhythmic production of REMs are perhaps at the origin of certain differences between our results and those in the literature (Krynicki 1975; Lavie 1979). Our results showed that REMs occurring during REM sleep were modulated by a certain number of statistically significant rhythmic frequencies. The periods corresponding to these frequencies ranged from 1 to 21 min, and no predominant rhythmic periods were present. As in the case of the slow trend, the existence of a given rhythmic period was not dependent on subject, night, time of night or length of the REM sleep phase. Of course we must keep in mind that the mathematical detection of a rhythmic period will depend on the length of the REM sleep phase being considered, in that we cannot test for a long rhythmic period in a very short REM phase. In conclusion, it is possible to verify statistically the existence of a rhythmic production of REMs in the majority of REM sleep phases. However, from one REM phase to another, the period of such rhythmic patterns may vary from 1 to more than 20 rain.
REMs D U R I N G REM SLEEP IN MAN
Summary The purpose of this study was to verify statistically the existence of certain characteristics of rapid eye movement (REM) production during REM sleep: organization and rhythmicity. REM data were collected intermittently, over a period of 20 consecutive nights, from 6 normal adult subjects of both sexes. Results concerning the organization of REM production revealed the presence of 3 evolving slow trends: quadratic, linear and quasi-sinusoidal. The occurrence of a given trend type was not related to subject, night, time of night or length of the REM sleep phase being analysed. Results of the analysis showed that it is possible for a series of REMs to contain one, two or more statistically significant rhythmic periods, or no significant rhythmic period. The rhythmic periods that were detected ranged from 1 to 21 min, and no predominant periods were present. As in the case of the slow trends, the occurrence of a given rhythmic period was not dependent on subject, night, time of night or length of the REM sleep phase.
R~sum~ Caractbristiques de survenue des mouvements oculaires rapides en sommeil paradoxal en r h o m m e : organisation et rythmicitb
L'objectif de cette 6tude est de v6rifier statistiquement l'existence de certaines caract6ristiques de survenue des mouvements oculaires rapides en sommeil paradoxal, h savoir leur organisation temporelle et leur rythmicit6. Les mouvements oculaires rapides ont 6t~ dbnombr~s lors de certaines nuits chez 6 personnes adultes des deux sexes ayant dormi pendant 20 nuits cons6cutives au laboratoire. En ce qui concerne l'organisation des mouvements oculaires rapides, les r6sultats montrent qu'il existe 3 sortes de tendances lentes: quadratique, lin6aire et quasi-sinusoidale. L'existence d'une ten-
155
dance lente donn6e ne d6pend pas du sujet, de la nuit, du moment de la nuit ou de la dur6e de la phase de sommeil paradoxal analys6e. Les analyses de rythmicit6 de la production des mouvements oculaires rapides montrent qu'il est possible de trouver, un, deux ou m~me plusieurs rythmes statistiquement significatifs, ou au contraire aucune rythmicit6 statistiquement confirm6e pour une m~me suite de mouvements oculaires. Les p6riodes correspondant aux rythmicit6s confirm6es statistiquement varient entre 1 et 21 rain, sans qu'il y ait pr6dominance d'une p6riode particuli6re. Comme dans le cas de l'organisation temporelle, l'existence d'une rythmicit6 particuli6re dans la production des mouvements oculaires rapides ne d6pend pas du sujet, de la nuit, du moment de la nuit, ou de la dur6e de la phase de sommeil paradoxal consid~r6e.
References Aserinsky, E. The maximal capacity for sleep: rapid eye movement density as an index of sleep satiety. Biol. Psychiat., 1969, 1: 147-159. Aserinsky, E. Rapid eye movement density and pattern in the sleep of normal young adults. Psychophysiology, 1971, 8: 161-175. Aserinsky, E. Relationship of rapid eye movement density to the prior accumulation of sleep and wakefulness. Psychophysiology, 1973, 10: 545-558. Blackman, R.B. and Tukey, J.V. The Measurement of Power Spectra. Dover Publications, New York, 1958. Krynicki, V. Time trends and periodic cycles in REM sleep eye movements. Electroenceph. clin. Neurophysiol., 1975, 39: 507 -513. Lavie, P. Rapid eye movements in REM sleep - - more evidence for a periodic organization. Electroenceph. clin. Neurophysiol., 1979, 46: 683-688. Rechtschaffen, A. and Kales, A. A Manual of Standardized Terminology, Techniques and Scoring Systems for Sleep Stages of Human Subjects. U.S. Government Printing Office (Public Health Service), Washington, D.C., 1968. Salzarulo, P. Variations with time of the quantity of eye movements during fast sleep in man. Electroenceph. clin. Neurophysiol., 1972, 32: 409-416. Zimmerman, J.C., Czeisler, C.A., Laximinarayan, S., Knauer, R.S. and Weitzman, E.D. REM density is dissociated from REM sleep timing during free-running sleep episodes. Sleep, 1980, 2: 409-415.
151
CHARACTERISTICS OF RAPID EYE MOVEMENT PRODUCTION DURING REM SLEEP IN MAN: ORGANIZATION AND RHYTHMICITY L.D. WEBER, A. MUZET, J.P. SCHIEBER and J.P. LIENHARD Centre d'Etudes Bioclimatiques du C.N.R.S., 21 rue Becquerel, 67087 Strasbourg C E D E X (France)
(Accepted for publication: October 1, 1982)
Recently, researchers have become interested in certain characteristics of rapid eye movement (REM) production during REM sleep in man. Several studies have been published concerning REM organization within a REM sleep phase. For example, Aserinsky (1971) found that in REM sleep phases equal to or greater than 20 min, ocular motor activity peaked 5-10 rain after the onset of REM sleep and then decreased 10 min later. Salzarulo (1972) reported that REM density variations showed a peak value, the location of which within the phase differed according to the length of the phase. He further stated that this density peak was reached more quickly in longer phases of REM sleep and that after 21 min of REM sleep, REM density always decreased progressively and regularly to reach a value near zero. Others have studied the rhythmic or periodic aspect of REM production. Krynicki (1975) found a rhythmic production of REMs whose period ranged from 10 to 30 min. Lavie (1979), however, found that this periodicity was limited to between 10 and 20 min. The purpose of the present study was to verify an organized and rhythmic production of REMs by means of a mathematical technique which would allow us to test individual spectral peaks that might be present in series of REMs.
Method
Six young adults, of both sexes, between 19 and 24 years of age, spent 20 consecutive nights in the sleep laboratory. During this time these subjects took part in a study concerning the effects of noise
on the sleep of different age groups. The first 3 nights were reference nights, the following 15 nights were disturbed by traffic noises and the last 2 nights were recovery nights. Lights out was at 23.00 h and subjects remained in bed until 07.00 h the following morning. During the reference nights (N1, N2), the two first and last disturbed nights (N3, N4, N16, N17) and the recovery nights (NI8, N I9), the following uninterrupted electrophysiological records were obtained: EEG (derivations F3-A2 and C3-A2); 2 electro-oculograms (EOGs; the electrodes were placed slightly above and below the outer canthus); EMG of either the mentalis or the masseter; ECG; finger pulse amplitude; pulse wave velocity; body mobility, measured by a bed actograph and a radar detection system. During the remaining noise-disturbed nights (N5N15), only cardiovascular records were made. The noise emitted during the disturbed nights of this study consisted of 3 traffic noises whose intensities were 45, 55 or 65 dBA and whose durations were 4, 8 and 12 sec respectively. These noises were continuously diffused during the night according to a semi-random temporal distribution at a rate of 90 noises/h. Other nocturnal ambient conditions were maintained constant for all experimental nights (air temperature of 20°C _+ I°C; a relative humidity of 60% _+ 5%; a background noise due to air conditioning of 35 dBA). All records were visually scored for sleep stages in 30 sec epochs, according to the criteria given by Rechtschaffen and Kales (1968). REMs were detected visually and counted by periods of 30 sec. Only REMs that were visible in both EOGs and with an amplitude of at least 5 mm (60 ~tV) in one of the two channels were taken into consideration
0013-4649/83/0000-0000/$03.00 © 1983 Elsevier Scientific Publishers Ireland. Ltd.
152
L.D. WEBER ET AL.
for this study. REM sleep phases which were less than 20 min in length, interrupted by more than 5 min of N R E M sleep or interrupted by experimenter intervention were not analysed. In all, a total of 114 REM sleep phases which occurred during the 3 reference nights, disturbed nights N3, N4, N16 and N17, and the 2 recovery nights (nights for which we had EOG records) qualified for statistical analysis.
Statistical analysis Estimation and subtraction of slow trends of REMs The method which was used consisted of adjusting a least squared polynome to our series of REMs. This was achieved by calculating the coefficients of 8~ for i = 0, 1 ... K of a polynome of K degree (Pi = 80 + ~ l X i -at-t~2 x 2 " ' " + ~ K X K ) • The polynomial values which were calculated for each i were then subtracted point by point from the observed series of REMs (formation of the differences: ~ ( i = X i - P i ) . The curve which was traced from the Pi polynomial values indicates the evolving tendency or slow trend of the observed series. The )(i values were then used to find possible periodic fluctuations in comparison with the slow trend. Periodicity or quasi-periodicity detection by a power spectral method The method of Fourier transformation of the autocorrelation function and its associated errors of estimation has been described by Blackman and Tukey (1958). The power spectrum which is obtained in this way must be corrected by adequate smoothing of the calculated spectral values (p.S.xi p. S.x, : p.s. = power spectrum; - indicates smoothing). Under these conditions p.'s. can be considered as an acceptable estimate of the real power spectrum of the series X i, or as a distribution of the total variance of the observed series in independent fractions, in a certain number of frequency bands. This property (Parseval theorem) is important because it allows for the possibility of testing for the existence of periodicity (significant variance) in one or more given spectral bands. ~
Thus, the independent fractions of the variance calculated for the different spectral bands can be compared among themselves either individually or cumulatively by comparing and weighting them with the associated degrees of freedom (Snedecor F). With this method, the longest rhythmic periods that can be detected are limited to N x 2, while the shortest periods are equal to 2 x the sampling interval (Shannon rule). This method was used to test the estimated values of variance in certain frequency bands where the variance appeared to be significantly higher than the 'residual' variance located in the remaining frequency bands of the spectrum. An example of this analysis can be found in Table I.
TABLE I Example of a spectrum of variance showing 2 zones (boxes) of cumulated spectral values that were visually detected and then tested for their statistical significance (Snedecor F test). The first spectral value was always eliminated, due to the accumulation of a large part of the variance in this part of the spectrum which corresponds to low frequencies. Frequencies are expressed in c/min. Frequency
Spectral value
0.0556 0.1111 0.1667 0.2222 0.2778 0.3333 0.3889 0.4444 0.5000 0.5556 0.6111 0.6667 0.7222 0.7778 0.8333 0.8889 0.9444 1.0000
4.0772
** P < 0.01.
1.1765 [~.392~ 12.3885 I 1.2401 0.8177 /0.7216 0.4453 0,5861 0.8954 0.8791 0.6932 0.5860 0.8463 1.0538 0.9901
F
df
1.39
16/127
2.33
16/127
S
**
R E M s D U R I N G REM SLEEP IN M A N
153
Results
60
Of the 114 REM sleep phases that were analysed: 11 were 1st REM sleep phases of the night, 34 were 2nd, 33 were 3rd, 29 were 4th and 7 were 5th.
--
B
50
40
~~~~
30
Slow trend analysis Results showed that there were 3 types of predominant REM production trends in our series of REMs. These evolving trends, along with their frequency of occurrence and the mean duration of the REM sleep phase during which they occurred, can be found in Table II. The occurrence of a given slow trend was not related to subject, experimental night or time of night. In addition, the slow trend form was not related to the length of the REM sleep phase, with the exception of those eye movement series that were shown to have no significant trend. In this case REM sleep phases were short in duration. However, in 20 REM phases that were shorter than 25 min, only 6 had no significant slow trends while the remaining 14 did. Detection of periodicity of rapid eye movement production The spectral analysis of REMs was made after having subtracted the slow trends of REMs. Afterwards, during both non-disturbed and disturbed nights, all dominant spectral values were visually detected and noted. Finally, these spectral values were tested by means of a Snedecor F test in order to expose those spectral values which were statistically significant. Results showed that during the
T A B L E II Types of slow evolving trends present in rapid eye movement series equal to or longer than 20 rain. Slow trend type
Frequency of occurrence
Per cent
Mean duration (min)
Quadratic Quasi-sinusoidal Linear Other No trend
52 28 20 8 6
46 25 17 7 5
40.9 37.3 42.6 39.2 21.6
visuaLLy detected rhythmic peaks N:269 statisticaLLy sigmficant rhythmic periods N:111
20 10 0
5
10
15
20 m;n
Length of rhythmic periods
Fig. 1. Histograms of spectral estimations (in rain). A distinction has been made between spectral peaks that were visually detected and spectral peaks which were statistically significant (Snedecor F test). Non-disturbed and disturbed nights have been cumulated.
non-disturbed nights, only 40% of the spectral values that were visually detected were in fact statistically significant. This proportion was 41% for the disturbed nights. During the non-disturbed nights, REMs were modulated by statistically significant ( P < 0.01) rhythmic periods which were inferior or equal to 3 rain in 14% of the cases, ranging between 3 and 6 rain in 33% of the cases, ranging from 6 to 9 min in 18% of the cases and superior to 9 min in 35% of the cases. During the disturbed nights these proportions were: 18%, 35%, 8% and 39% respectively. Because of the large similarity between the distribution histograms of the rhythmic peaks during the non-disturbed nights and disturbed nights, we cumulated our observations in order to increase the quantity of our data. The corresponding histogram can be found in Fig. 1.
Discussion
Our results are in agreement with some of the conclusions reached by Aserinsky (1971), Krynicki (1975) and Lavie (1979), in that they support the
154
notion of an organized as opposed to a random distribution of REMs during REM sleep. Slow trend detection in our series revealed several predominant trend types. Like others (Salzarulo 1972; Krynicki 1975) we found quadratic and linear type trends, in addition to quasi-sinusoidal type trends. The occurrence of a given trend was not dependent on subject, night, time of night or length of the REM sleep phase. We found, as did Krynicki (1975), that at least one of the quadratic trends presented a concave-up form rather than the more common concave-down form, and that at least two of the linear trends were increasing rather than decreasing. The explanation of the presence of these different production trends in our series of REMs is for the moment speculative. Nonetheless, in the light of our findings and a certain number of those in the literature, we are led to the following hypothesis. It is possible that these different slow trends may be in actuality the expression of parts of the same, continuous oscillating process controlling ocular motor activity, this process being present throughout the night. If we admit such an underlying process, we should then propose that REM sleep phases would occur during different parts of this process, thus accounting for the presence of different slow trend types. This hypothesis would suggest that while the appearance of REMs during REM sleep would be controlled by a REM sleepdependent mechanism, the characteristics of the production of REMs (trend type and rhythmic period) would be controlled by a mechanism of another nature, being independent of REM sleep. Our hypothesis supports the findings of Zimmerman et al. (1980), who found that REM density was dissociated from REM sleep timing under free-running sleep conditions. Their results showed that while there existed differences in REM sleep accumulation and duration in entrained and free-running sleep conditions, there were no such differences in REM production patterns. Further, our hypothesis supports those made by Aserinsky (1969, 1973) in which he related increases in REM density during the night to the amount of prior accumulated sleep. Before discussing the results of the detection of rhythmic or periodic production of REMs during
L.D. WEBER ET AL.
REM sleep, we would like to specify what we think are two important points concerning the method used in this analysis. Firstly, certain authors chose not to subtract the slow trend from REM series before the spectral analysis of these series. The inconvenience of this method is the accumulation of a large part of the variance in that part of the spectrum which corresponds to low frequencies. The origin of this 'energy' (variance) is not necessarily of a harmonic nature. In addition, the presence of this energy, often relatively large in relationship to the rest of the spectrum, makes the detection of rhythms by analysis of variance less sensitive. For these reasons we found it preferable to eliminate the slow trend. Secondly, it is possible that a series of REMs may contain one, two or more statistically significant rhythmic periods, or no significant rhythmic period. Thus, we found it necessary to test all visually detected spectral peaks for their statistical significance. As can be seen from our results, approximately 60% of the spectral peaks which were visually detected in our series of REMs were not statistically significant. These two points concerning the analytical method which we used in detecting the rhythmic production of REMs are perhaps at the origin of certain differences between our results and those in the literature (Krynicki 1975; Lavie 1979). Our results showed that REMs occurring during REM sleep were modulated by a certain number of statistically significant rhythmic frequencies. The periods corresponding to these frequencies ranged from 1 to 21 min, and no predominant rhythmic periods were present. As in the case of the slow trend, the existence of a given rhythmic period was not dependent on subject, night, time of night or length of the REM sleep phase. Of course we must keep in mind that the mathematical detection of a rhythmic period will depend on the length of the REM sleep phase being considered, in that we cannot test for a long rhythmic period in a very short REM phase. In conclusion, it is possible to verify statistically the existence of a rhythmic production of REMs in the majority of REM sleep phases. However, from one REM phase to another, the period of such rhythmic patterns may vary from 1 to more than 20 rain.
REMs D U R I N G REM SLEEP IN MAN
Summary The purpose of this study was to verify statistically the existence of certain characteristics of rapid eye movement (REM) production during REM sleep: organization and rhythmicity. REM data were collected intermittently, over a period of 20 consecutive nights, from 6 normal adult subjects of both sexes. Results concerning the organization of REM production revealed the presence of 3 evolving slow trends: quadratic, linear and quasi-sinusoidal. The occurrence of a given trend type was not related to subject, night, time of night or length of the REM sleep phase being analysed. Results of the analysis showed that it is possible for a series of REMs to contain one, two or more statistically significant rhythmic periods, or no significant rhythmic period. The rhythmic periods that were detected ranged from 1 to 21 min, and no predominant periods were present. As in the case of the slow trends, the occurrence of a given rhythmic period was not dependent on subject, night, time of night or length of the REM sleep phase.
R~sum~ Caractbristiques de survenue des mouvements oculaires rapides en sommeil paradoxal en r h o m m e : organisation et rythmicitb
L'objectif de cette 6tude est de v6rifier statistiquement l'existence de certaines caract6ristiques de survenue des mouvements oculaires rapides en sommeil paradoxal, h savoir leur organisation temporelle et leur rythmicit6. Les mouvements oculaires rapides ont 6t~ dbnombr~s lors de certaines nuits chez 6 personnes adultes des deux sexes ayant dormi pendant 20 nuits cons6cutives au laboratoire. En ce qui concerne l'organisation des mouvements oculaires rapides, les r6sultats montrent qu'il existe 3 sortes de tendances lentes: quadratique, lin6aire et quasi-sinusoidale. L'existence d'une ten-
155
dance lente donn6e ne d6pend pas du sujet, de la nuit, du moment de la nuit ou de la dur6e de la phase de sommeil paradoxal analys6e. Les analyses de rythmicit6 de la production des mouvements oculaires rapides montrent qu'il est possible de trouver, un, deux ou m~me plusieurs rythmes statistiquement significatifs, ou au contraire aucune rythmicit6 statistiquement confirm6e pour une m~me suite de mouvements oculaires. Les p6riodes correspondant aux rythmicit6s confirm6es statistiquement varient entre 1 et 21 rain, sans qu'il y ait pr6dominance d'une p6riode particuli6re. Comme dans le cas de l'organisation temporelle, l'existence d'une rythmicit6 particuli6re dans la production des mouvements oculaires rapides ne d6pend pas du sujet, de la nuit, du moment de la nuit, ou de la dur6e de la phase de sommeil paradoxal consid~r6e.
References Aserinsky, E. The maximal capacity for sleep: rapid eye movement density as an index of sleep satiety. Biol. Psychiat., 1969, 1: 147-159. Aserinsky, E. Rapid eye movement density and pattern in the sleep of normal young adults. Psychophysiology, 1971, 8: 161-175. Aserinsky, E. Relationship of rapid eye movement density to the prior accumulation of sleep and wakefulness. Psychophysiology, 1973, 10: 545-558. Blackman, R.B. and Tukey, J.V. The Measurement of Power Spectra. Dover Publications, New York, 1958. Krynicki, V. Time trends and periodic cycles in REM sleep eye movements. Electroenceph. clin. Neurophysiol., 1975, 39: 507 -513. Lavie, P. Rapid eye movements in REM sleep - - more evidence for a periodic organization. Electroenceph. clin. Neurophysiol., 1979, 46: 683-688. Rechtschaffen, A. and Kales, A. A Manual of Standardized Terminology, Techniques and Scoring Systems for Sleep Stages of Human Subjects. U.S. Government Printing Office (Public Health Service), Washington, D.C., 1968. Salzarulo, P. Variations with time of the quantity of eye movements during fast sleep in man. Electroenceph. clin. Neurophysiol., 1972, 32: 409-416. Zimmerman, J.C., Czeisler, C.A., Laximinarayan, S., Knauer, R.S. and Weitzman, E.D. REM density is dissociated from REM sleep timing during free-running sleep episodes. Sleep, 1980, 2: 409-415.