INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

It has been known for some time that low-level muscle activity occurs during sleep in association with slow wave components in the electroencephalographic (EEG) signal and that this muscle activity is inhibited during rapid-eye-movement (REM) sleep (cf. Pompeiano 1967; Steriade and McCarley 1990 for reviews). It is also recognized that patients with pain of a putative musculoskeletal origin, where muscle activity may be part of the pain induction process, often suffer from poor sleep (e.g., fibromyalgia, chronic fatigue syndrome, myofascial pain syndrome) (Goldenberg 1993; McCain and Scudds 1988). We therefore conducted a study with surface electromyographic (EMG) recordings from the trapezius muscles to investigate possible association between EMG activity during sleep and sleep quality, including subjectively experienced overnight discomfort or pain. In a few supplementary recordings, upper extremity muscles were included. The study is an extension of earlier work in this field and included also daytime monitoring of EMG and heart rate (Jensen et al. 1993a; Vasseljen and Westgaard 1995; Westgaard et al. 2001).

Inspection of the EMG recordings confirmed that periods with sustained low-level muscle activity were present during sleep. The root-mean-square (RMS) value of the detected EMG activity showed a characteristic pattern, featuring periods where the EMG signal amplitude was remarkably stable over periods lasting as long as 1 h. There was little short-term variation in the EMG data, with the striking exception of the consistent appearance of low-frequency amplitude oscillations at frequencies ranging from ~0.05 to 0.2 Hz. Although rhythmic muscle activity has been illustrated in many studies pertaining to sleep physiology (e.g., Arduini et al. 1963; Glenn and Dement 1981; Marchiafava and Pompeiano 1964), to our knowledge this activity has not been properly analyzed before. The purpose of this paper is to present this phenomenon, describe its characteristics, and discuss possible neurophysiological mechanisms in the generation of this muscle activity pattern.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects and data recording

Twenty-seven female subjects (mean age 45.1 yr, range 31-62 yr) participated in the study following written informed consent. The subjects were recruited from a health-care center with elderly people (health-care personnel without heavy physical work tasks) and from a shopping center (retail personnel). The purpose of the study, as explained to the subjects, was to further the insight in work-related shoulder pain by comparison of muscle activity during work and after working hours. About half of the subjects had experienced moderate shoulder pain during the last 6 mo but were still working, the other half reported no shoulder pain. No one had diagnosed sleep disorders. During daytime (~9 AM to 5 PM), the working hours were spent in retail customer service or providing health care for patients that in most cases were physically able to provide for themselves. Outside working hours the subjects carried out their normal activities without receiving any instructions as to preferred or restricted activities. They were not instructed with respect to the use of caffeine or other stimulants. Another five females, recruited from the same subject base, participated in supplementary experiments with simultaneous recordings of trapezius and upper extremity muscles.

Electrocardiographic (EKG) and surface EMG signals were continuously recorded over a 24-h period. A lightweight, battery-powered data monitor equipment (Physiometer, Premed) with data storage provided by a flash card mounted in a pocket computer (HP 200LX, Hewlett-Packard) was used as the recording device. Electrodes for recording the EKG signal were placed in standard positions across the chest. The surface representation of the cardiac potential with Q, R, and S peaks (the QRS complex; Malik and Camm 1995) was detected. The intervals between the R peaks (RR intervals) were derived on a beat-by-beat basis. The surface EMG activity from left and right trapezius muscles was monitored by bipolar electrode pairs and included a ground electrode (10- to 800-Hz bandwidth). The EMG electrodes (circular with diameter 6 mm, Blue Sensor, E-10-vs, Medicotest A/S) were placed on a line from the C₇ vertebra to the acromion process. The medial electrode was placed 2.5 cm lateral to the midpoint of the line and the inter-electrode distance was 20 mm. This electrode placement was shown to produce a stable EMG signal even with electrode movement in relation to the underlying muscle (Jensen et al. 1993b). The ground electrode was placed on the spine of the C₇ vertebra. The EMG recording device was used to process the raw signals by calculating the RMS value using a 0.1-s time window. EKG and EMG data were stored on the HP computer at a sampling rate of 10 Hz. On transfer of data to a stationary PC, the time resolution was reduced to 0.2 s by averaging adjacent data points.

Before the ambulatory recording, a maximal voluntary contraction (MVC) in 90° arm abduction was performed twice by the subject to acquire the maximal EMG amplitude. This was used as a reference value to calibrate the EMG response during the ambulatory recordings (% EMG_max) (Jensen et al. 1996). Daytime results are not reported in this paper; however, the trapezius muscle activity was generally low in the awake state with the median EMG activity level for most subjects in the range 2-5% EMG_max (Westgaard et al. 2001). The duration of the sleep period varied from 4.7 to 8.4 h, mean 6.8 h. Onset of sleep was determined in these recordings by a sudden change in the EMG activity pattern from irregular, phasic activity to no activity, or presenting a dramatically changed pattern with near-constant EMG amplitude and minimal temporal variation (cf. Fig. 1).

View larger version (58K): [in this window] [in a new window]   Fig. 1. Electromyographic (EMG) pattern observed during sleep. A: time course of the root-mean-square (RMS) value of the EMG signal recorded from the left upper trapezius muscle during sleep. Two arrows on the bottom of the plot indicate the time when the subject fell asleep (sleep on) and woke up (sleep off) respectively. Several intervals of almost constant EMG amplitude are shown in a range corresponding to values between 1 and 2% EMGmax. Three arrows and associated shaded areas marked by S1, S2, and S3 indicate 3 10-min time intervals selected to illustrate the observed phenomenon more in details (see bottom). B: the RMS-detected EMG data. C: the corresponding frequency content. For low-level EMG activity (bottom left) the square of the magnitude of the Fourier transform (on the right side) shows a narrowband component at approximately 1 Hz due to the EKG interference. For higher EMG activity levels, signal components were generally detected in the 0.05- to 0.2-Hz range.

In five subjects, the upper trapezius EMG recordings were supplemented by EMG recordings from lower trapezius, deltoid, biceps, and forearm flexors (flexor carpi ulnaris, flexor digitorum superficialis). The bipolar electrode pairs were placed on the muscle bellies, and the EMG signal calibrated by the EMG response in resisted MVCs with attempted movement in the muscle force direction. In these recordings, a strain gauge transducer placed around the chest monitored respiratory activity.

The quality of the EMG signal was checked by visual inspection at the beginning and the end of the recording and by observing the activity pattern at rest and during a maximal voluntary contraction. During the recording, an automated quality evaluation procedure checked for artifacts by detecting slow swings in baseline and/or sharp transients in the raw signal. Intervals detected by this procedure were marked and the RMS amplitude examined off-line. Such events were rare, <0.1% of the data points, unless gross failures due to breakage of leads or unfastening of electrodes occurred. Of the 27 subjects, successful sleep recordings were obtained from right and/or left trapezius in 44 cases, 22 from each side. Both recordings failed in four subjects, one recording failed in two subjects. In the successful recordings, the RMS amplitude of data points marked by the artifact detection procedure rarely deviated significantly from neighboring data points, and corrective measures were therefore not carried out. The artifact detection procedure did not respond to interference from the heartbeat, which could be particularly evident in recordings from the left upper trapezius during periods with low EMG activity. This noise component in the EMG signal was dramatically attenuated through digital filtering in the off-line data processing (cf. next section).

Data processing

The time series constituted by the RMS values of the EMG signal computed over 0.2-s time intervals was first low-pass filtered to attenuate possible EKG interference. A Chebyshev (type II) low-pass filter was designed with transition band 0.70-0.75 Hz, 40 dB of attenuation in the stopband, and 1 dB maximum loss in the passband. Using a bilinear transformation (with prewarping), the corresponding infinite impulse response (IIR) filter was implemented and applied to the RMS time series as a noncausal filter (i.e., after filtering in the forward direction, the series was reversed and filtered again).

A series of processing procedures were carried out to characterize the observed oscillatory phenomenon that appeared to affect the EMG amplitude. To measure the amplitude of the rhythmic components that marked the RMS time series, we first selected time intervals in which the rhythmic components were present and then computed the SD of the time series over 512-sample epochs with 50% overlap. The estimated values were plotted as a function of the mean RMS value of the EMG for each epoch.

A second step of the procedure was aimed at characterizing the RMS time series in the frequency domain. The square of the magnitude of the Fourier transform of time intervals of the RMS time series that showed a rhythmic component was estimated to assess the frequency content of the analyzed signal. In addition, because of the nonstationarity of the RMS time series, we applied a Choi-Williams transformation ( = 1) and thus derived the time-frequency distribution (Williams 1997) of the RMS time series.

To compare the frequency content of the RR time series with that of the RMS time series, the frequency analysis procedure was also applied to the RR time series. Both the square of the magnitude of the Fourier transform of the signal as well as its Choi-Williams time-frequency representation ( = 1) were computed and compared with the EMG data. For the recordings including respiratory activity, time-frequency analysis was applied also to the recorded output of the strain gauge transducer.

Additional to the processing of the sleep recordings, the firing pattern of single motor units (MUs) recorded from awake subjects in an experimental situation with mental stress and minimal body movement was examined. This data set was recorded in an earlier study to observe motor unit substitution (Westgaard and De Luca 1999). The modulation of firing pattern in these recordings was investigated by an autoregressive technique. The frequency content of MU firings was obtained using the Burg method and the order of the model was chosen using the Akaike criterion (Ljung and Glad 1994).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The RMS amplitude of the EMG recording exceeded 1% EMG_max for 10 min in 15 (8 from right, 7 from left trapezius) of 44 successful recordings. Each of these recordings showed an obvious rhythmic component, and the results of the analysis of these 15 recordings are presented here. There was no apparent association between the EMG response and the shoulder pain reported by the subjects.

Figures 1 and 2A illustrate the processing of one recording. The typical appearance of RMS-detected muscle activity during sleep is shown in Fig. 1A. Long periods of elevated, near-invariant EMG activity are interspersed between periods without discernable activity and short bursts of phasic activity, presumably due to spontaneous shifts in posture.

View larger version (29K): [in this window] [in a new window]   Fig. 2. Relationship between RMS value of the EMG rhythmic component and amplitude of the EMG data. A: the results for the recording illustrated in Fig. 1. A linear relationship is shown between the amplitude of the rhythmic component that marks the RMS time series and the actual amplitude of the EMG signal. , data before filtering; , the results after applying a low-pass filter to attenuate possible interference from the electrocardiographic (EKG) signal. The regression lines for these 2 sets of data are also shown. B: the regression lines for all the recordings with responses higher than 1% EMGmax performed in this study. The data range is shown by the length of the regression lines with the exception of 1 case where EMG amplitudes up to 8.5% EMGmax were detected. This plot indicates that for most of the recordings, but a few exceptions, the slope of the regression line was approximately the same. See text for details.

Inspection of periods with elevated EMG activity at higher time resolution reveals low-frequency oscillatory behavior at frequencies <0.2 Hz as outlined by the square of the magnitude of the Fourier transform of the time series. This is illustrated in Fig. 1B where three epochs selected from the RMS time series presented in Fig. 1A are represented (marked S1-S3). These three epochs have different mean values and frequency contents as pointed out by the square of the magnitude of their Fourier transform (normalized) shown in the lower panels (Fig. 1, B and C). A trend of larger amplitude oscillations with increasing mean EMG amplitude is evident. At low activity levels the EMG signal was in some recordings contaminated by the EKG signal, detected as a distinct peak at ~1 Hz in the square of the magnitude of the Fourier transform of the RMS time series (see Fig. 1, B and C, bottom). A low-pass digital filter (as described in METHODS) was used to attenuate this noise component.

Figure 2A shows the scatter plots of RMS oscillatory amplitude versus mean RMS amplitude before and after the application of the low-pass filtering procedure for the recording shown in Fig. 1. The data points are selected from segments with very low EMG activity and from segments with relatively stable RMS amplitude, up to ~2% EMG_max. The effect of filtering is most evident at low amplitudes, as anticipated. A further effect of the procedure is to reduce the scatter of the individual data points from the regression line (R² = 0.91 unfiltered, R² = 0.98 after filtering). However, the increase in RMS oscillatory amplitude with increasing RMS mean amplitude is clear in both the filtered and unfiltered data.

All recordings with EMG responses higher than the noise level showed evidence of the oscillatory behavior illustrated in Fig. 1. Regression lines of RMS oscillatory amplitude versus mean RMS amplitude were constructed after filtering, as shown for one subject in Fig. 2A. The regression lines for all 15 recordings that reached amplitudes >1% EMG_max are represented in Fig. 2B. The length of the lines indicates the data range. With the exception of one recording, the regression lines showed a near-constant association (i.e., slope) between oscillatory amplitude and EMG amplitude. The mean slope coefficient is 0.017 ± 0.007 (SD), corresponding to approximately a 10% ratio of the peak-to-peak oscillatory amplitude to mean EMG amplitude. The deviation of data points from the regression line was low for all subjects; the mean value of R² was 0.87 ± 0.11.

The low frequency of the EMG oscillations may suggest an influence from the sympathetic nervous system as the source of the motoneuron excitation, by analogy with research on the heart rate variability (e.g., Malliani et al. 1994; Task Force of the European Society of Cardiology and the North American Society of Pacing and Electrophysiology 1996). Rhythmic activity of sympathetic origin, known from the analysis of heart rate variability, includes control of circulation (0.1 Hz) and temperature regulation (<0.05 Hz). Respiratory activity in quiet breathing typically has frequencies centered ~0.25 Hz. We therefore compared rhythmic activity in muscle with simultaneously recorded heart rate, by examining the square of the magnitude of the Fourier transform as well as the time-frequency distribution of the RMS and the RR time series. The comparison showed that different frequency components contribute to EMG and RR time series.

An example of the analysis is shown in Figs. 3 and 4, presenting time plots of simultaneous EMG amplitude and RR time series over a 10-min period, the square of the magnitude of their Fourier transform, and the contour plot of their time-frequency distribution. It is clear that the EMG and the RR time series are constituted by different frequency components. Both signals show a nonstationary behavior. The time-frequency representation of the EMG time series shows a single component ~0.15 Hz that is modulated in frequency in the range between 0.1 and 0.2 Hz. The time-frequency representation of the RR time series shows a quasi-stationary signal component ~0.2 Hz and two components <0.1 Hz associated with signal nonstationarities. There is no apparent coupling in the dynamic progression of frequency variation between the two responses. This observation was confirmed by other recordings.

View larger version (43K): [in this window] [in a new window]   Fig. 3. Example of RMS time-series analysis. Bottom: the time course of the signal. Right: represents the (normalized) square of the magnitude of the Fourier transform of the RMS time series. Top left: the contour plot of the time-frequency distribution of the RMS time series. The contour plot is obtained by displaying five contour levels at 10, 30, 50, 70, and 90% of the maximum of the distribution. The time-frequency representation shows that the oscillations associated with the EMG AM are constituted by a single narrowband component, which is strongly frequency modulated over time.

View larger version (43K): [in this window] [in a new window]   Fig. 4. Example of heartbeat time series analysis. Bottom: time series analysis of the intervals between the R peaks (RR time series) in the surface representation of the heartbeat potential (Malik and Camm 1995). Top right: the square of the magnitude of the Fourier transform of the signal. Top, left: relative to the time-frequency distribution of the RR time series. The time-frequency representation indicates that the RR series is a multi-component signal associated with nonstationarities particularly evident at low frequencies. The recording was performed simultaneously to the RMS time series presented in Fig. 3. The comparison of Figs. 3 and 4 indicates that these 2 time series are constituted by different frequency components. Also, there is no apparent relationship between the evolution in time of the RMS and RR time series.

A similar comparison of EMG responses was made for the few examples of simultaneous responses (>1% EMG_max) in both trapezius muscles. An example is shown in Fig. 5 where plots of the RMS time series of right and left trapezius are shown in the bottom panels while their time-frequency representation is presented in the corresponding top panels. The frequency variations of the EMG AM in the two muscles appear to follow each other. It should be noted that while the FM of the EMG amplitude on right and left upper trapezius follows basically the same pattern, the mean amplitude of the EMG signal as well as the peak-to-peak amplitude of the oscillations is clearly different.

View larger version (58K): [in this window] [in a new window]   Fig. 5. Simultaneous recording of right and left upper trapezius EMG activity. Bottom: the time course of the RMS time series recorded from left (A) and right (B) upper trapezius. Top (A and B): the corresponding time-frequency distributions. It is indicated that although the average amplitude of the EMG signal as well as the associated oscillations are of different magnitude, the time-frequency content of the 2 signals follows the same pattern.

The RMS time series was occasionally constituted by two narrowband components rather than by a single component as shown in the previous example. One of these observations when two components were detected in the RMS time series is reported in Fig. 6. The bottom plot reports the time course of the signal, whereas the top plot represents its time-frequency distribution. Two components are shown with similar (but not always equal) FM.

View larger version (53K): [in this window] [in a new window]   Fig. 6. Example of a RMS time series constituted by 2 distinct frequency components. Bottom: the time course of the signal. Top: time-frequency distribution. The 2 frequency components appear to be associated with similar but not always equal FM.

The two phenomena of multiple frequency components and same EMG oscillatory behavior of different muscles during sleep are further illustrated in an experiment with simultaneous recordings from upper and lower trapezius, deltoid, biceps, and forearm flexors. Figure 7 shows a 10-min section of recording with EMG oscillatory modulation in these muscles as well as heart rate and respiratory rhythm. A strong oscillatory rhythm ~0.2 Hz was observed in biceps and forearm flexors (Fig. 7, D and E). Lower trapezius presented the same rhythm in a weaker response (Fig. 7C) and the rhythm is observed in recordings from upper trapezius and deltoid (Fig. 7, A and B). However, in these last two muscles, a rhythm at about half the frequency is the more dominant one, presenting a two-component activity pattern in deltoid, whereas a three-component pattern (including respiratory modulation; Fig. 7G) is apparent in upper trapezius. The muscle activity patterns are not coherent with the modulation of heart rate, with the exception of the respiratory related rhythmic component, which is dominant in the RR time series (Fig. 7F) and is also present in the upper trapezius RMS time series (Fig. 7A).

View larger version (33K): [in this window] [in a new window]   Fig. 7. Time-frequency analysis of a 10-min section of rhythmic EMG activity recorded from upper trapezius (A), deltoid (B), lower trapezius (C), biceps (D), and forearm flexor muscles (E). Heart rate (F) and respiration (G) were simultaneously recorded. All muscles show a similar, rhythmic modulation centered around a frequency of 0.2 Hz. In addition, upper trapezius and deltoid show a rhythmic frequency component of ~0.1 Hz and the upper trapezius shows also respiratory modulation. Respiratory modulation is clearly represented in the RR time series.

Finally, evidence of a rhythmic component in the same frequency range of the above-described observations from surface EMG recordings was found on investigation of trapezius MU activity pattern recorded in a situation with mental stress (Fig. 8). The MU firing pattern was more irregular (noisy) than the surface EMG time series recorded during sleep thus preventing us from performing time-frequency analysis of the MU activity; however, by use of an autoregressive technique, two distinct peaks in the firing rate modulation were found, centered around 0.25 and 0.1 Hz, respectively. The higher frequency peak presumably represents respiratory modulation, while the FM ~0.1 Hz may point to activation by the same premotor pathways that cause muscle activity modulation during sleep.

View larger version (22K): [in this window] [in a new window]   Fig. 8. Frequency content (top) of the firing pattern of trapezius MU (bottom) in an awake subject during mental stress. Frequency analysis by an autoregressive technique demonstrates 2 dominant peaks, a presumed respiratory component at 0.25 Hz and a component of unknown origin at 0.1 Hz. The latter component is in the same frequency range as the observations from surface EMG data recorded during sleep.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The phenomenon of low-FM of muscle activity pattern during sleep was an unanticipated observation made during an experimental series carried out for different purposes. Important features of the phenomenon are a constant ratio of oscillatory amplitude to mean EMG amplitude, similar rhythmic activity in the contralateral trapezius muscles, and the rhythmic activity corresponds to frequencies typically associated with sympathetically generated activity but is different from the sympathetic rhythms manifest in the heart rate variability. We have further demonstrated that muscle activity simultaneously recorded from upper and lower trapezius muscles and muscles of the upper extremities may deviate in their time-frequency patterns and that rhythmic MU activity with a similar frequency content may exist in the awake state, at least in an experimental situation with mental stress.

Before discussing possible origins of the observed phenomenon, we will point out that raw EMG signals were stored and later examined in some of the later recordings. We can thus rule out the occurrence of possible artifacts not necessarily evident from the RMS time series. Also because a continuous AM is apparent in these raw EMG recordings, the observed phenomenon may be not attributed to obvious causes (e.g., periodic limb movement disorder) that would result in periodic bursts of EMG activity.

The time intervals of relatively constant EMG amplitude are preceded and followed by abrupt changes in the EMG signal. We suggest that these abrupt changes are related to postural adjustments. In this case, we expect a displacement between the surface electrodes and the muscle fibers from which the EMG data are recorded; however, our previous work suggests that this displacement is unlikely to affect the EMG amplitude at the chosen electrode location for trapezius (Jensen et al. 1993b).

Recent developments in sleep research describe physiological phenomena that may be associated with the rhythmic component of modulation of the muscle activity reported in this study. Because we documented that the FM of the observed rhythmic component is consistent for right and left trapezius muscles and for muscles further separated, we discuss in the following "central" mechanisms that could be related to this phenomenon. We have, however, no explanation for the phenomenon of dual frequency components appearing in some, but not all muscles.

The presence of widespread slow waves (<1 Hz) in the cortical electroencephalogram has been reported (Steriade et al. 1993a-c). These waves are synchronized over wide cortical regions, including the pyramidal neurons in the motor cortex (Amzica and Steriade 1995; Contreras and Steriade 1997), and pyramidal tract neurons present a bursting activity pattern during sleep (Evarts 1964). The frequency content of the cortical waves is higher than the muscle activity oscillations documented in this study, i.e., >0.3 Hz in cats (Steriade et al. 1993b,c) and 0.5-1.0 Hz in humans (Achermann and Borbély 1997). However, the slower envelopes that describe the prevalence of slow wave activity and the pattern of occurrence of spindle activity present a frequency range similar to the rhythmic component of muscle activity documented in this study (Achermann and Borbély 1997).

The change in frequency content of the modulation of EMG amplitude during sleep may reflect a changing balance between slow wave and spindle activity, as the envelopes of these EEG patterns appear to have different frequencies of dominance. In the example illustrated in Fig. 5 of Achermann and Borbély (1997), the first two sleep episodes are dominated by slow wave activity centered around 0.05 Hz, the last two episodes are dominated by spindle activity at ~0.23 Hz. The envelopes of rhythmic activity in cortical EEG are associated with similar fluctuations in the firing of cells in the midbrain and brain stem reticular formation (Oakson and Steriade 1983; Trulson and Jacobs 1979).

The muscle activity pattern observed during sleep shows striking similarities (low-level, invariant activity patterns) to muscle activity observed in laboratory experiments with induced stress (Wærsted et al. 1996; Westgaard and Bjørklund 1987; Westgaard and De Luca 1999). The surface EMG signal in these experiments does not usually include an oscillatory component, however, we here show that the firing rate of individual motor units may be modulated at frequencies similar to those observed during sleep. The oscillations in the firing rate of individual motor units when performing tasks with induced stress are typically 1-2 pulses per second (pps), corresponding to ~10% of the mean firing rate. This is in good agreement with the ratio of EMG amplitude oscillation to mean EMG amplitude observed during sleep. If all motor units participated in synchronized firing, a rhythmic component in the AM of the surface EMG should be observable. This behavior is rarely observed, possibly because many premotor pathways without oscillatory patterning contribute to the excitation of motoneurons in the awake state.

The observation of low-frequency, rhythmic MU firing when awake, if confirmed in more extended analyses, may implicate mechanisms and pathways other than cortical rhythmic activity influencing motoneuron firing through cortico-spinal pathways. One alternative is activity in premotor monoaminergic pathways from the brain stem. These include serotonergic pathways from the raphe nuclei (Holstege 1991; Jacobs and Fornal 1997) and noradrenergic pathways from locus coeruleus (Jacobs et al. 1991). The activity of serotonergic raphe neurons is generally described as regular or "clock-like" firing at 3-4 Hz. Slow modulation of this activity at frequencies similar to those observed in this study is reported (Montagne-Clavel et al. 1995). The envelope of this FM may cause AM of the muscle activity pattern with a frequency content as herein documented. Different activity patterns within the raphe nuclei (Trulson and Trulson 1982) or between different brain stem structures that show rhythmic activity patterns (Lydic et al. 1983) may provide clues as to the pathways involved in the motoneuron excitation.

It may be considered surprising that the rhythmic muscle activity, which by frequency content is under influence of the sympathetic nervous system, is not associated with similar frequency bands in the heart rate variability. However, the slow components of the heart rate rhythm are related to specific control functions, such as respiration and circulation. These control functions may not be prominently displayed in the CNS structures that are a putative source of the observed motoneuron excitation during sleep.

Our observation of rhythmic muscle activity may provide an important window for observation of physiological processes during sleep, by representing the CNS output to an important effector organ. The motor output allows a clear description of dynamic activity components by the time-frequency analysis utilized in our data processing procedure, and may provide useful information to supplement, e.g., EEG-recorded activity patterns.