See discussions, stats, and author profiles for this publication at:

https://www.researchgate.net/publication/13533171

Scalp recorded direct current brain potentials

during human sleep

Article

in

European Journal of Neuroscience · April 1998

Impact Factor: 3.18 · DOI: 10.1046/j.1460-9568.1998.00131.x · Source: PubMed

CITATIONS

30

READS

107

4 authors

, including:

Lisa Marshall

Universität zu Lübeck

82

PUBLICATIONS

4,365

CITATIONS

SEE PROFILE

Matthias Mölle

Universität zu Lübeck

79

PUBLICATIONS

5,094

CITATIONS

SEE PROFILE

Available from: Matthias Mölle

Retrieved on: 18 July 2016

 

L. Marshall et al.    DC potential shifts during nocturnal sleep               1167 

European Journal of Neuroscience

 

Vol. 10. pp. 1167-1178, 1998 

 

Scalp recorded direct current brain potentials during  
human sleep

 

 
Lisa Marshall,

1

, Matthias Mölle,

1

, Horst L. Fehm

2

 and & Jan Born

1,3

 

1

 Department of Clinical Neuroendocrinology, Medical University of Lübeck, 23538 Lübeck, Germany,  

2

Department of Internal Medicine, Medical University of Lübeck, 23538 Lübeck, Germany, 

3

Department of Physiological Psychology, University of Bamberg, 96045 Bamberg, Germany  

 

Abstract 
 
The direct current (DC) potential recorded from the scalp of awake humans has been considered a reflection of general 
changes in cortical excitability. This study examined DC potential shifts in humans during a night of continuous sleep. 
Standard polysomnographic recordings and skin temperature were measured simultaneously. Contrary to expectations, 
average DC potential level indicated higher negativity during nonrapid eye movement (NREM) sleep than REM sleep and 
wakefulness. Moreover, a dynamic regulation of the DC potential level was revealed in association with the NREM-REM 
sleep cycle comprising four successive phases: (i) a steep  NREM transition negative shift  during the initial 10-15 min of 
the NREM sleep period; (ii) a more subtle  NREM positive slope  during the subsequent NREM sleep period; (iii) a steep 

REM transition positive shift  starting shortly prior to the REM sleep period, and (iv) a  REM negative slope , 

characterizing the remaining greater part of the REM sleep period. DC potential changes were only weakly related to 
changes in slow wave activity (r 

2

< 0.18). The NREM negative slope and REM positive slope could reflect, respectively, 

gradually increasing and decreasing cortical excitability resulting from widespread changes in the depolarization of apical 
dendrites. In contrast, the NREM transition negative shift and the REM transition positive shift may reflect the 
progression and retrogression, respectively, of a long lasting hyperpolarization in deeply lying neurons. 
 
Correspondence: L. Marshall, as abouve. E-mail: marshall@kfg.mu-luebeck.de 
 
Introduction 

 

Direct current (DC) potentials recorded from the cortical 
surface or scalp are considered primarily to be neuronal in 
origin (e.g. Caspers et al. 1984). Approximations based on 
electrodynamic principles led to the assumption that 
widespread cortical activity is essential in producing large 
amplitude surface potentials in humans (Birbaumer  et al
1990; Elbert 1993). It has become established to associate 
negative DC or slow potential shifts with widespread 
synchronized membrane depolarizations of pyramidal 
apical dendrites. The electrogenesis of positive shifts is 
less firmly established, and may merely result from a 
reduction in surface negativity. However, field potentials 
generated by sources in deeper cortical layers, i.e. 
widespread hyperpolarization in these layers, may also 
produce a superficial potential of negative polarity (Mitzdorf 
1985;  Creutzfeldt 1995). Shifts of the DC potential may 
also reflect modifications in the recently described slow 
oscillations (< 1 Hz) emerging in association with 
synchronized sleep (Steriade  et al. 1993a) and/or in the 
infraslow potential oscillations (T 

8 s, and 

1 min, 

Aladjalova 1964).  

According to one of the few comprehensive theories, 

proposed by the group of Birbaumer and Elbert, slow 
cortical DC potential changes during wakefulness can be 
described as a measure reflecting a tonic tuning of 
excitability of cortical neuronal networks, apparently only 
loosely linked to the faster, oscillatory activity of the 
electroencephalogram (EEG), and the transitory pro-
cessing of discrete elements of incoming information 
(Elbert & Rockstroh 1987;  Birbaumer  et al. 1990; for 
related references:  Born  et al. 1982;  Stamm  et al. 1987; 
McCallum 1993). A distinct regulation of excitability within 
widespread cortical neuronal networks presumably also 
occurs in association with sleep, possibly serving functions 
of cortical information processing specifically related to this 
state of consciousness, e.g. the formation of memory 
(Tilley  et al. 1992;  Cipolli 1995;  Sejnowski 1995;  Buszáki 
1996).  

Distinct DC potential shifts  in association with different 

states of sleep and wakefulness have been described in 
several studies in various animal species (cp.  Caspers 
1993). DC potential recordings have usually been found to 
shift toward positive values at the transition from 
wakefulness to sleep, with increasing positivity occurring 
with the development of slow wave sleep; negative going 
shifts from wakefulness to slow wave sleep have, 
however, also been reported (cp.  Bechtereva 1974 for a 
review). In humans, at the transition from wakefulness to 
sleep, a negative DC displacement has been found 
repeatedly (Hoffmann  et al. 1988;  Marshall  et al. 1994, 
Marshall et al. 1996a). The recording of DC potentials 
from the human scalp throughout the whole nocturnal 
sleep period has not been thoroughly documented as yet 
(Davis  et al. 1939). This derives from difficulties in 
avoiding biological and technical artifacts, e.g. a nonstable 
recording system, potential drifts related to fluctuations in 
electrode/electrolyte equilibrium, temperature, blood gas 
tension, and to drifts related to skin potentials, or 
movements of the subject (Butler 1993;  Caspers 1993). 
However, provided a thorough control of these artifact 
sources, extensive investigations have yielded strong 
evidence that DC potentials recorded from the scalp 
reflect essentially cortical activity (e.g.  Rockstroh & 
McCallum 1993).  

The aim of the present study was to measure DC 

potential shifts during the entire period of normal nocturnal 
sleep in humans. Distinct potential levels in relation to the 
sleep wake activity, as well as to the NREM-REM sleep 
cycle were awaited. Based on findings in animals, it was 
hypothesized, (i) that the sleep period would reveal a 
mean potential level on average more positive than during 
wakefulness, coherent with a reduced level of cortical 
excitability during sleep, and (ii) that REM sleep would be 
associated with transient increases in superficial 
negativity. Preliminary findings of this experiment have 
been reported previously in abstract form (Marshall  et al
1996b).

 

 

L. Marshall et al.    DC potential shifts during nocturnal sleep               1168 

Materials and methods 
 
Subjects and procedure
  
Eight subjects (four men, four women) aged 25.6 ± 0.9 
years (mean ± SEM) voluntarily participated in two night 
sessions, i.e. one night to accustom subjects to the 
laboratory and one experimental night. Before the 
adaptation night, subjects had practised at home during 
three consecutive nights to sleep throughout the night lying 
on their back while wearing an orthopaedic cervical collar 
(Philadelphia, Brevete, Canada), which restrained the head 
so that its movement in relation to the body was prevented. 
Volunteers reporting not being able to sleep for at least 5 h 
in such a fashion at home were excluded from the study 
before participating in the adaptation night. By such 
screening, eight subjects had been immediately discarded. 
Procedures in the adaptation night were identical with 
those in the experimental night, except in the former, 
electrodes for DC potential recording were not applied and 
subjects arrived at the laboratory 2 h later. Application of 
electrodes for DC potential recording during the adaptation 
night was deemed unnecessary since once applied, 
subjects cannot discriminate between electrodes for 
regular EEG and DC potential recording. Subjects had no 
history of sleep disturbances, nor did they take any 
medication at the time of the experiments. They usually 
went to bed before 24 h and got up between 06:00 and 
07:30 h. They had been instructed to go to sleep at 23:00 
h, to get up at 06:30 h, not to take a nap on the day of the 
experimental session, and to restrain from caffeinated and 
alcoholic drinks in the afternoon of the experimental day. 
The studies took place in a sleep laboratory at the Medical 
University of Lübeck, and the experimental protocol was 
approved by the local ethics committee. 

On the experimental day, the subject arrived at the 

laboratory at 19:00 h to prepare recordings of (i) DC 
potentials, (ii) skin temperature and (iii) standard 
polysomnographical measures. At 22:00 h the subject was 
assigned to bed. The cervical collar (used during home 
practice) was fitted around the subject s neck and the 
subject was to position him /herself comfortably in a supine 
position, whereby head and neck rested in the head
shaped cavity of a foam rubber cushion (Bisanz, Gau
Algesheim, Germany). Subsequently, the cervical collar 
was fastened onto the cushion, thus, suppressing head 
movement. Prior investigations had shown it essential to 
restrict especially head movements during nocturnal DC 
potential recordings to exclude artefacts caused by 
movement and external temperature changes at the 
recording locations. For this reason, subjects were to 
maintain their supine position throughout the night, with 
electrode locations limited to frontal and central cortical 
areas. 

When lights were turned off at 23:00 h, the subject was 

asked to close his/her eyes, but to maintain wakefulness 
while lying quietly over the next 10-15 min. To  control 
wakefulness and induce a defined mental state during this 
period, the subject was to keep mental count of time and 
gently press a hand held button every estimated 30 s. 
Withdrawal of the button press device signalled to the 
subject to try to maintain wakefulness for a few more 
minutes, and then succumb to sleep. Subjects had been 
instructed to lie quietly and avoid any gross movements 
throughout the whole period of experimental recordings. At 
06:30 h, a dim light was turned on and subjects gently 
awakened. They were again given the button press device 
in order to control wakefulness and induce a stable mental 
state for another 10-15 min. Since subjects and their 
polysomnographic recordings were monitored, if subjects 
awoke spontaneously after 06:00 h they were given the 

button press device at that time. In the adaptation night, 
the procedure was the same as in the experimental night, 
with the exception of the DC potential measurement. 
 
Recording  
DC potential recordings were obtained from electrodes 
located at F3, F4, C3, C4 (International 10 20 6System, 
(Jasper 1958), FC3 and FC4. Locations FC3 and FC4 
were positioned halfway between F3 and C3, and F4 and 
C4, respectively. All electrodes were referenced to linked 
electrodes attached at the mastoids. The ground 
electrode was positioned in the centre of the forehead. In 
applying the electrodes, care was taken to avoid possible 
contamination from skin potentials (Picton & Hillyard 
1972). After rubbing the scalp site thoroughly with alcohol, 
a  clip on  electrode socket (Bauer  et al. 1989) was 
attached with collodion, and the scalp punctured with a 
sterile hypodermic needle until minor bleeding occurred. 
Electrode gel (Electrode Electrolyte, TECA Corp., NY, 
USA) was applied to the socket, and the electrode, also 
filled with gel, was clipped on. Non polarizable Ag/AgCl 
electrodes (8 mm diameter, ZAK GmbH, Simbach/Inn, 
Germany) were used. Air bubbles had been previously 
eliminated from the electrode gel by centrifugation (1300 
U/min for 3 h at 10 °C) and care was taken not to produce 
any air bubbles when clipping electrodes onto their 
sockets. Electrodes, as well as electrode gel for DC 
potential recordings, were kept and applied in the same 
room and at the same ambient temperature as that under 
recording conditions. Prior to use, DC electrodes had 
been connected pairwise via electrode gel for at least 7 h 
to reduce any electrode bias potential (Girton &  Kamiya 
1974). Electrodes for DC recordings were attached at 
least 1 h prior to the beginning of experimental recordings, 
a time sufficient to allow for the skin electrode interface to 
stabilize. Electrode impedance was measured 
immediately after application of all electrodes as well as at 
the very end of the recording session, and was usually 
below 2 kOhm, and always below 5 kOhm.  

A direct current amplifier (Toennies DC/AC amplifier, 

Jaeger GmbH & Co. KG, Würzburg, Germany, input 
resistance 2  100 MOhm) was used for recording and 
amplification of DC potentials. With short circuited input, 
the amplifier drifts, if present, were below 3  V/h. The 
amplification was set at 1000 fold, the low pass filter at 10 
Hz, and the threshold for automatic DC offset correction at 
± 4 mV. Analog DC signals were digitized (CED 1401, 
Cambridge Electronic Design Ltd, UK) at a rate of 50 Hz 
per channel and stored on a PC together with a time 
marker (every 30 s) for off line analysis. 

Skin temperature was measured every 32 s from three 

temperature probes, one near each mastoid reference 
and one on the scalp in the vicinity of C3 or C4. 
Temperature data were stored by a digital Mini logger (± 
0.1 °C level of accuracy, Mini Mitter, Sunriver, OR, USA). 
Derivations for standard polysomnographic sleep 
recordings of EEG (Fz, Cz), electrooculogram, and 
submental electromyogram were obtained using 
nonpolarizable Ag/AgCl electrodes (diameter 8 mm, 
Sensormedics, The Netherlands), signals were amplified 
by a Nicolet electroencephalograph, and written together 
with the 30 s time marker on paper for polysomnographic 
scoring. 
 
Data processing and statistical analysis
  
To adapt all data to a common mean time scale, average 
values of DC potentials and skin temperature were 
calculated for 30 s epochs. Correspondingly, sleep 
stages (1, 2, 3, 4, REM sleep, and wake time) and 
movement artefacts were scored off line for 30 s 

 

L. Marshall et al.    DC potential shifts during nocturnal sleep               1169 

intervals, according to the criterion of  (Rechtschaffen & 
Kales 1968), using the standard polysomnographical sleep 
recordings.  

With regard to DC potentials, resettings due to the 

amplifier s automatic DC offset correction were removed 
by adding the offset value to the successive potential 
values. Epochs of EEG revealing movement times as 
defined by (Rechtschaffen & Kales 1968) occurred utmost 
seldom (three times), and any transient DC potential shift 
occurring simultaneously was eliminated. Two approaches 
were chosen to examine the time course of the DC 
potential: (i)  The time course was examined for the total 
night, composed of a short epoch of wakefulness in the 
evening, the long interval of nocturnal sleep, and a final 
short epoch of morning wakefulness. (ii) The time course of 
the DC potential was also analysed with reference to 
successive NREM-REM sleep cycles during sleep. For 
calculating grand means of the DC potential, data points of 
all subjects were transformed to a common mean time 
scale based on the mean durations of the respective 
epochs of wakefulness and sleep time (when the total night 
was examined), or of NREM and REM sleep (when the 
relation to the NREM-REM sleep cycle was examined). For 
analysis of the course of the DC potential during the total 
night, the average potential during the initial epoch of 
wakefulness was set to 0 

V. Analyses were then 

conducted for original DC potential values, but also after 
linear trends had been removed, separately, for each 
electrode and subject (Hennighausen  et al. 1993). 
Analyses of the DC potential with reference to the NREM-
REM sleep cycle were  performed after linear trends 
occurring across the corresponding epoch had been 
removed from the DC potential data. In these latter 
analyses, the first 10 data points for a certain time interval 
of interest were set to 0  V.  

Based on classical sleep scoring according to 

(Rechtschaffen & Kales 1968), for each experimental night, 
sleep onset latency (with reference to the time when the 
button press device was withdrawn), total sleep time, time 
and percentage of time spent in the different sleep stages 
and in wakefulness were determined. In addition, latencies 
of sleep stages with reference to sleep onset were 
determined. The beginning of sleep was defined as the 
time when a subject passed from wakefulness to sleep 
stage 1 followed by a period of stage 2 sleep, without 
intermediate awakening. NREM and REM sleep periods 
and their completion were defined according to (Feinberg & 
Floyd 1979), i.e. besides a 15 min minimum for NREM 
sleep periods, the last NREM sleep period of the night was 
considered complete if it was followed by 5 min or longer of 
REM sleep before awakening. Similarly, the last REM 
sleep period was considered complete if followed by 5 min 
or longer of NREM sleep. During the night, episodes of 
REM sleep were considered to constitute a REM sleep 
period if at least 5 min in duration, except for the first REM 
sleep period, for which no minimum length was stipulated. 
A REM sleep episode interrupted by less than 5 min of 
continuous NREM sleep or wakefulness was treated as a 
single REM sleep period. Likewise, a NREM sleep episode 
interrupted by less than 5 min of continuous REM sleep or 
wakefulness was treated as a single NREM sleep period if, 
as mentioned above, at least 15 min in duration. The 
number and duration of awakenings were calculated for the 
entire interval of sleep, and for the NREM and REM sleep 
periods. Average durations of the NREM and REM sleep 
periods, which included the above mentioned short 
interruptions, were also determined (cp. legends to Figs 4, 
6 and 7).  

For a more fine grained analysis of the relation between 

DC potential changes and NREM sleep, slow wave activity 

(SWA) was also determined. SWA (i.e. power density 
within the delta frequency band) represents a continuous 
measure of NREM sleep intensity (Achermann & Borbély 
1987), and in contrast to the discrete sleep stages, lends 
itself to correlation with the DC potential data. An off line 
Fast Fourier Transform was used on the DC potential 
recordings to calculate power density in the 0.49-4 Hz 
band. One value was obtained for each 30 s epoch, 
having been computed from five overlapping windows of 
10.24 s (512 samples) each. Data from each of the five 
windows had been multiplied by a raised cosine function 
before the transform in order to taper the signal towards 
zero at the extremes of the data window, thus reducing 
errors induced by edge effects.  

Statistical analysis was generally based on analyses of 

variance followed by contrasts to specify significant 
effects and interaction terms. With regard to the DC 
potential (and also SWA), analyses of variance included 
repeated measures factors representing topography (F3, 
F4, FC3, FC4, C3, C4) and the different time intervals to 
be compared. The topographical distribution of potential 
changes was also tested by introducing a factor 
representing the frontal to central direction (F, FC, C). For 
the latter analyses, average values were formed between 
F3 and F4, FC3 and FC4, and C3 and C4 to obtain F, FC 
and C, respectively. 

The time courses of DC potential shifts and SWA for 

each electrode location were statistically compared by 
cross correlation analysis using time lags up to ± 45 min 
(i.e. 181 time points). Cross correlation coefficients were 
first calculated separately for each subject and then 
transformed with Fisher s Z transformation. Mean 
coefficients  r 

c

 were then tested for significance using 

Student s  t  tests. Sequential time lags with P< 0.05 are 
indicated in corresponding diagrams by a horizontal line 
and asterisk. Corresponding cross correlation analyses 
were performed to investigate a possible influence of 
temperature changes on the time course of the DC 
potential. All values in the text are given as mean ± SEM.

 

 
 
Results
 

 
Sleep parameters are shown  in  Table 1. Mean sleep 

time and time spent in the different sleep stages were 
very close to typical sleep under laboratory conditions (cp. 
Hirshkowitz  et al. 1992). Sleep onset latency and time 
spent in sleep stage 1 appeared to be increased, which 
was probably due to the restrained position of head and 
body during sleep. Sleep fragmentation, in general, was 
not a problem, as revealed by a normal mean number of 
awakenings per night: 5.6 (range 0-15), with a mean 
duration of 2.6 min (0-6.5 min). The number of 
awakenings per NREM sleep period, and the number of 
awakenings and short interruptions per REM sleep period 
both averaged 0.8/period (range 0-5), with a mean 
duration of these interruptions of 1.8 min (range 0.5-4.5). 
Mean duration of the NREM and REM sleep periods 
averaged 80.2 ± 7.6 min and 13.8 ± 4.5 min for the first, 
and 81.8 ± 5.2 min and 23.7 ± 4.0 min for the second to 
fourth NREM and REM sleep cycles. 
 
Time course of the DC potential during a total night s 
sleep
  
Fig. 1reveals the averaged time courses of the DC 
potential shifts (before and after elimination of linear 
trends) at frontal and central recording locations, and of 
SWA. Fig. 2shows respective recordings from two 
individual nights.  Before elimination of linear trends, the 
averaged DC potential during sleep time tended to be  

 

L. Marshall et al.    DC potential shifts during nocturnal sleep               1170 

Table 1. Mean time (± SEM) spent in different sleep stages and mean sleep latencies (± SEM) as defined by Rechtschaffen & Kales 1968)  

 

Time 

Mean  (SEM) [min] 

Mean  (SEM) [min] 

Latencies 

Mean  (SEM) [min] 

 

 

 

 

 

 

 

 

13.3  (5.4) 

3.1  (1.3) 

Sleep onset 

18.6  (2.0) 

S1 

53.8  (8.5) 

12.5  (2.0) 

S2 

6.8  (2.1) 

S2 

191.7  (6.9) 

44.3  (2.3) 

SWS 

32.2  (12.7) 

S3 

43.0  (4.8) 

9.9  (1.0) 

REM 

92.8  (20.7) 

S4 

54.8  (12.4) 

12.3  (2.8) 

 

   

SWS 

97.8  (12.9) 

22.2  (2.6) 

 

   

REM 

78.3  (7.5) 

17.9  (1.5) 

 

   

Total sleep time 

434.8  (9.2) 

   

 

   

 

Left: Total sleep time, and time spent in the different sleep stages and in wakefulness in minutes, as well as in percentage of total sleep time. Right: 
Sleep onset latency with reference to withdrawal of the button press device, and latencies of different sleep stages with reference to sleep onset. 
Wakefulness (W), sleep stage 1(S1), 2(S2), 3(S3), 4(S4); slow wave sleep (SWS) defined by the sum of sleep stages 3 and 4, REM (rapid eye 
movement) sleep (REM). n=8. 

 

 

more negative at frontal (F, mean of F3 and F4) than 
central (C, mean of C3 and C4) recording locations (F 

1,7

 = 

3.79;P< 0.1). The DC potential at fronto central (FC, 
halfway between frontal and central) and central locations 
shifted toward positivity during the recording time, i.e. from 
evening wakefulness, across sleep, to morning wake-
fulness (F 

3,21

 = 9.06;P< 0.05, for the effect of time 

averaged across FC and C; Fig. 1).  

The elimination of linear trends (  Fig. 1B), clarified the 

distribution of the sleep associated DC potentials. The 
effect of time concentrated on recordings over frontal sites 
(F

3,21

 =4.76;P< 0.05). Here, average DC negativity during 

total sleep time (-143.1± 74.3 V) was higher than during 
morning wakefulness (+82.4± 85.0 V;P<0.01, for a pair-
wise comparison). Frontocortical DC potential negativity 
during sleep, particularly over the right hemisphere, was 
also higher than during evening wakefulness (set to 0 
V;P< 0.05). Average DC potential levels between the first 
and second half of sleep time did not differ at any electrode 
site. During sleep time, but not during morning wake-
fulness, the mean DC potential level at frontal electrode 
locations was more negative (F: -143.0 ± 90.2  V) than at 
fronto central (FC: +43.5 ± 114.5  V), or central locations 
(C:-43.4±103.7  V;P<0.05, for each pairwise comparison).  

SWA was higher during the first (1.82 ± 0.14 mV

2

/Hz) 

than the second half of sleep time (1.41 ± 0.14 mV

2

/Hz), 

and at a minimum during evening (0.66 ± 0.06 mV

2

/Hz) and 

morning (0.59 ± 0.04 mV

2

/Hz) wakefulness (F 

3,21

 = 

37.92;P< 0.001, for effect of time,  P< 0.05, for pairwise 
comparisons between any two of the time intervals except 
between the two intervals of wakefulness). During sleep, 
but not during the periods of wakefulness, SWA was more 
pronounced at frontal (F: 1.10 ± 0.12 mV

2

/Hz) than fronto

central recording locations (FC: 0.94 ± 0.11 mV

2

/Hz), and 

at the latter locations activity was still higher than at central 
recording locations (C: 0.81 ± 0.10 mV

2

/Hz;F 

2,14

 = 

67.08;P< 0.001, for effect of topography, and P< 0.001 for 
pairwise comparisons between any two locations).  

Since topographical differences in the DC potential level 

as well as in SWA basically only occurred in the frontal to 
central direction, i.e. between the frontal and central 
locations, comparisons between the time course of these 
measures by cross correlation functions will be restricted 
to these electrode locations. Cross correlation coefficients 
between DC potential and SWA were generally low. 
Nevertheless, some of the correlation coefficients reached 
significance (Fig. 3): cortical DC potential shifts (after 
elimination of linear trends) anticipated changes in SWA by 
10-20 min indicating that DC potential shifts toward 
positivity were followed by decreases in SWA. However, 
average correlations at these time lags were still below r 

c

 = 

0.24, i.e. variances of the DC potential changes explained 
by prior changes in SWA remained below 6%. Also, cross
correlations revealed significance for SWA anticipating DC 
potential shifts by 20-45 min indicating that greater SWA 
was followed by DC potential shifts directed toward more 

positive values. Again, respective correlation coefficients 
were rather low and did not exceed r 

c

 = 0.27.  

Changes in skin temperature remained under 0.5 °C 

throughout all nights, except in two instances when a shift 
of 0.7 and 1.3 °C at one of the recording sites occurred. 
Associated DC shifts were transient and artefact conta-
mination could be excluded. Cross correlations between 
mean DC potential levels, after elimination of linear 
trends, and skin temperature were always less than r 

c

 = 

0.12 and non significant. 
 
Transitions between wakefulness and sleep
  
An analysis restricted to 15 min periods prior to and after 
sleep onset was performed to examine changes in the DC 
potential (after elimination of linear trends) at the transition 
from wakefulness to sleep. As compared to the preceding 
15 min mean level of wakefulness, the DC potential levels 
showed a negative shift across all electrode locations (F 

3,21

 = 7.13;P< 0.01).  Table 2 summarizes the results of 

this analysis. The shift at frontal locations was again 
larger, i.e. more negative, than at fronto central (F 

1,7

 = 

4.17;P< 0.08) and central recording sites (F 

1,7

 = 17.35;P

0.01). Although the negative DC potential shift upon sleep 
onset at frontal and fronto central locations appeared to 
be somewhat more consistent over the right than left 
hemisphere (  Table 2), the corresponding analyses of 
variance main effect failed to reach significance (F 

1,7

 = 

3.39;P< 0.11).  

A corresponding analysis restricted to the 15 min period 

prior to and after the time of awakening revealed a signi-
ficant change in the DC potential level toward positivity 
during wakefulness, across all electrode sites (F 

3,21

 = 

6.01;P< 0.05). There was, however, no indication of any 
topographic difference in the DC potential during morning 
wakefulness (F 

5,35

 = 1.12;P> 0.3). 

 
The time course of the DC potential during the NREM-
REM sleep cycle
  
Fig. 4shows the average sleep profile and the mean time 
course of the DC potential and SWA for the first four sleep 
cycles of six subjects. The other two subjects did not 
exhibit sufficiently long periods of NREM and REM sleep 
in succession. Comparing NREM and REM sleep, the DC 
potential at frontal and central locations was more positive 
during REM (average across F and C: 254.2 ± 51.5  V) 
than NREM sleep (65.5 ± 81.1  V;F 

1,5

 = 11.39, P< 0.05). 

Again, compared with the potential level at the central 
location, the DC potential at the frontal site was shifted 
towards negative values with this difference being equally 
consistent during NREM sleep (F 

1,5

 = 8.65;P< 0.05) and 

REM sleep (F 

1,5

 = 7.46;P< 0.05).  

The DC potential displayed a specific dynamic pattern 

during the NREM-REM sleep cycle. Although the 
prominent features of the temporal pattern of DC potential 
shifts (as shown in Fig. 5) were readily discernible during 
all four NREM-REM sleep cycles, resemblance appeared  

 

L. Marshall et al.    DC potential shifts during nocturnal sleep               1171 

0

60 120 180 240 300 360 420

mV

-0.4

0.0

0.4

0.8

1.2

0

60 120 180 240 300 360 420

mV

-0.3

-0.2

-0.1

0.0

0.1

0.2

min

0

60 120 180 240 300 360 420

mV

2

/Hz

0

1

2

3

4

EW

MW

A

B

C

S L E E P

central 

frontal 

Fig. 1Sleep across the total night. From top to bottom: 
averaged time courses of (A) DC (direct current) potential 
shifts before linear trend elimination, (B) DC potential shifts 
after linear trend elimination and, (C) slow wave activity 
across the total experimental night including a short period 
of evening wakefulness (EW, mean ± SEM: 22.3 ± 2.6 
min), the long interval of nocturnal sleep time (434.8 ± 9.2 
min), and a final period of morning wakefulness (MW, 14.8 
± 0.7 min). Thick lines represent averages of F3 and F4 
(frontal), and thin lines represent averages of C3 and C4 
(central). n = 8.  
 

to be greater among the second to fourth cycle, with the 

pattern of DC shifts during the first cycle exhibiting 
variations in some of its aspects. This led us to separately 
examine averaged time courses of the DC potential across 
the second to fourth and first NREM-REM sleep cycles. 
Figures 6 and  7 depict averaged time courses of sleep 
stage, DC potential shifts and SWA for the second to fourth 
and first NREM-REM-NREM sleep period, respectively. (In 
order to obtain a continuous course of the DC potential 
concurrent not only with the NREM to REM sleep, but also 
with the REM to NREM sleep transition, not merely NREM-
REM periods, but NREM-REM-NREM sleep periods served 
as bases for calculations.)  

Fig. 5 shows schematically the four main phases of the 

DC potential changes which could be extracted from all 
NREM-REM sleep cycles:  

1 the  NREM transition negative shift , which is a steep 
potential shift toward negativity during the initial 10-15 min 
of the NREM sleep period;  
2 the  NREM positive slope , which is a more subtle, 
gradual increase in positivity of the DC potential during 
the subsequent NREM period. Smaller DC potential shifts 
which appear to accompany intermittent shifts between 
stages of NREM sleep were often superimposed on this 
slope;  
3 the  REM transition positive shift , which is a steep 
positive potential shift reaching maximum values shortly 
after REM sleep onset and commencing prior to the 
polysomnographically defined REM sleep period;  
4 the  REM negative slope , which is a gradual potential 
shift toward negativity during the remaining REM sleep 
period, and can advance into the beginning of the 
subsequent NREM sleep period.  

For calculation of slopes of the discriminated DC 

potential phases, the  beginning and end of each phase 
were derived from the time course of the averaged curve 
and matched with individual records. Inspection of 
individual records indicated that, despite some variability 
concerning the positioning of the NREM transition
negative shift, the time intervals derived from average 
curves yielded reasonable estimates for the different 
phases of DC potential regulation. Table 3 gives the time 
intervals and corresponding slope values for the four 
phases of the DC potential changes for the second to 
fourth and for the first NREM-REM sleep cycles, 
separately. For the second to fourth NREM-REM cycle, all 
differences in the DC potential slope between any two 
subsequent phases of DC potential regulation were 
significant, whereas differences between slopes during 
the first NREM-REM cycle revealed modifications. Most 
prominent, the increase in positive voltage during the first 
NREM positive slope was much steeper than that during 
the second to fourth NREM-REM sleep cycle (F 

1,6

 = 

15.72;P< 0.01, Table 3). The other phases of DC potential 
regulation did not differ significantly between the first 
versus second to fourth NREM-REM sleep cycle, although 
for the REM transition positive shift (P< 0.2), a distinction 
was implied.  

SWA was, as expected, greater during NREM (2.11 ± 

0.13 mV

2

/Hz) than REM sleep periods (0.99 ± 0.04 

mV

2

/Hz;F 

1,5

 = 101.03;P< 0.001). Like the negativity of the 

DC potential, SWA was greater at frontal (1.64 ± 0.09 
mV

2

/Hz) than central recording locations  (1.46 ± 0.07 

mV

2

/Hz), with this effect being equally consistent during 

NREM sleep (F 

1,5

 = 80.70;P< 0.001) and REM sleep (F 

1,5

 = 76.79;P< 0.001). SWA was smaller during NREM 

sleep of the second to fourth than of the first sleep cycle 
(F 

1,5

 = 10.11;P< 0.05).  

Despite some similarities between the time courses of 

SWA and the DC potential during the NREM-REM sleep 
cycle, the course of the DC potential was unique in most 
aspects: while during the first part of NREM sleep the DC 
potential initially shifted negative and shortly following 
turned toward positivity, SWA consistently increased 
during this time (Fig. 4). During REM sleep, SWA 
remained essentially constant at a low level while the DC 
potential was mostly negative going.  

Cross correlation analyses between the DC potential 

shift and SWA conducted for the first as well as for the 
second to fourth NREM-REM sleep cycles revealed 
essentially the same pattern of results as the cross
correlation analyses obtained for the whole night, as 
depicted in Fig. 2. Somewhat higher correlation  

 

L. Marshall et al.    DC potential shifts during nocturnal sleep               1172 

1

2

3

4

5

6

7

8

9

-0.8

-0.4

0.0

0.4

0.8

1

2

3

4

5

6

7

8

9

0

1

2

3

4

5

6

7

1

2

3

4

5

6

7

8

9

4

3

2

1

REM

W

1

2

3

4

5

6

7

8

9

4

3

2

1

REM

W

1

2

3

4

5

6

7

8

9

-0.4

0.0

0.4

0.8

1

2

3

4

5

6

7

8

9

0

1

2

3

4

5

6

7

mV

mV

2

/Hz

II 

I

mV

2

/Hz

mV

hr

hr

A

B

C

frontal

central

 

Fig. 2Sleep across the night for two individual subjects (I, II). From top to bottom: time courses of (A) sleep stages, (B) DC 
potential shifts after linear trend elimination and, (C) slow wave activity, across the total experimental night including a 
short period of evening wakefulness and a final period of morning wakefulness. Wakefulness (W), sleep stage 1 (1), 2 (2), 
3 (3), 4 (4) and REM (rapid eye movement) sleep (REM). Thick lines represent averages of F3 and F4 (frontal), and thin 
lines represent averages of C3 and C4 (central).  

 

coefficients (peaking at r 

c

 = 0.44) were found for the first 

NREM-REM cycle, however, only at C3 and C4 for DC 
potential shifts anticipating decreases in SWA by about 20 
min. Corresponding analyses between DC potential levels 
and skin temperature did not reveal significant cross
correlations for any time lag. 
 

Discussion 
 
The present study describes the time course of scalp 
recorded DC potential shifts in human nocturnal sleep. To 
the authors  knowledge, this is the first time that in 
humans the dynamic changes of cortical DC potentials 
recorded continuously throughout a complete night of  

 

L. Marshall et al.    DC potential shifts during nocturnal sleep               1173 

F3

-45

-30

-15

15

30

45

-0.5

-0.3

0.3

0.5

*

*

lag [min]

r

c

F4

-45

-30

-15

15

30

45

-0.5

-0.3

0.3

0.5

*

lag [min]

r

c

C4

-45

-30

-15

15

30

45

-0.5

-0.3

0.3

0.5

*

*

lag [min]

r

c

C3

-45

-30

-15

15

30

45

-0.5

-0.3

0.3

0.5

*

*

lag [min]

r

c

 
Fig. 3

Cross correlation functions between cortical DC potential shifts and SWA (slow wave activity) across the total 
experimental night including a period of evening wakefulness, nocturnal sleep time and morning wakefulness at F3, F4, 
C3 and C4. Cross correlations analyses were conducted using time lags between -45 and +45 min. Time lags revealing 
significant (P< 0.05) cross correlations are marked by a horizontal bar and asterisk. Correlations (r 

c

) represent averages 

across eight nights.  

Table 2. The transition from wakefulness to sleep 

 

 

0-5 min 

Mean 

 
(SEM) 

5-10 min 

Mean 

 
(SEM) 

10-15 min 

Mean 

 
(SEM) 

 
F (3,21) 

 

 

 

 

 

 

 

 

F3 

-63.5  (52.0) 

-166.8  (49.4) 

-172.2  (64.9) 

5.99* 

F4 

-122.0  (41.7) 

-249.5  (54.0) 

-278.8  (72.0) 

12.59** 

FC3 

-47.6  (58.4) 

-132.6  (58.2) 

-139.0  (66.0) 

3.60 

FC4 

-50.6  (56.5) 

-158.4  (66.2) 

-183.8  (73.6) 

5.47* 

C3 

-60.6  (38.8) 

-142.3  (46.9) 

-157.2  (63.8) 

5.27* 

C4 

-41.8  (40.0) 

-124.0  (45.5) 

-138.5  (53.7) 

5.99* 

 

Mean shifts in DC (direct current)  potential (µV) during consecutive 5-min intervals following sleep onset as compared to the mean 
potential level during the 15-min interval of wakefulness preceding sleep (set to 0µV). Column on the right indicates F-values from 
ANOVA including repeated measures factors for the time intervals, for each electrode site separately. One and two asterisks indicate 
P<0.05 and P<0.01, respectively, for remaining F-values P<0.1. n=8.

 

 
 
 

sleep have been reported. DC potentials were recorded in 
conjunction with standard polysomnographic recordings, 
thus allowing for detailed sleep stage associated analysis. 
Contrary to our expectations, the average DC potential 
during sleep was more negative (over the frontal cortex) 
than during the adjacent phases of morning and evening 
wakefulness, and the DC potential was also more negative 
during NREM than REM sleep periods (Fig. 1). This finding 
contrasts with results from experiments of DC potential 
changes in wakefulness which suggested that negativity of 
the cortical DC potential is associated with increased 
cortical excitability deriving from widespread depolarization 
of apical dendrites (Birbaumer  et al. 1990). Rather than 
increased cortical excitability, the enhanced negative DC 
potential during sleep and NREM sleep could result from a 
long lasting hyperpolarization in deeper layers (Steriade et 
al
. 1993a). Even more important than the observed 
differences in the average DC potentials appears to be that 
neither during REM sleep nor during NREM sleep periods 
was there a steady DC potential level. Instead, the NREM-
REM sleep cycle as a whole was associated with a unique 
recurrent temporal pattern of changes in the DC potential 
suggesting a dynamic regulation of cortical activity (Fig. 6). 
The time course of the DC potential during the NREM-REM 

sleep cycle was comprised of four phases: a steep 
NREM transition negative shift , a gradual  NREM
positive slope , a steep  REM transition positive shift  
and a  REM negative slope  (Fig. 5).  

Long term DC potential recordings must be controlled for 
several sources of artefact (see Introduction). Drifts of the 
recording device in the short circuited condition were 
controlled intermittently and were several magnitudes 
smaller than the sleep associated shifts (see Materials 
and methods). Also, motor activity was excluded by 
fixating the subject s head during the recording epoch and 
excluding any epochs with clear muscle artefacts. With 
regard to DC potential changes linked to rapid eye 
movements in REM sleep, note that inspection of 
respective individual recording periods in this study, as in 
a former study (Marshall  et al. 1996a), revealed that DC 
potential shifts in association with phasic eye movements  

 

 

L. Marshall et al.    DC potential shifts during nocturnal sleep               1174 

0

60

120

180

240

300

360

mV

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

min

0

60

120

180

240

300

360

mV

2

/Hz

0

1

2

3

4

5

0

60

120

180

240

300

360

4

3

2

1

REM

W

A

B

C

central 

NREM 

1

REM 1

NREM 2

REM 2

NREM 

3

REM 3 NREM 

4

REM 4

frontal 

min

mV

REM

NREM

NREM

+

-

4

NREM-transition-negative shift

NREM-positive slope

REM-transition-positive shift

REM-negative slope

2

2

1

1

3

0

40

80

120

160

Fig. 4  Averaged time courses of sleep stage (A), DC 
potential shifts (B), and slow wave activity (C), for the first 
four sleep cycles. Thick lines represent averages of F3 and 
F4 (frontal), and thin lines represent averages of C3 and 
C4 (central). The durations of all NREM (nonrapid eye 
movement) and REM sleep periods 1-4 correspond to the 
average duration of each of these intervals. (Mean ± SEM 
duration in min were for NREM sleep periods 1-4, 
respectively, 80.3 ± 9.0, 82.0 ± 5.2, 87.3 ± 7.4, 64.2 ± 9.2, 
and for REM sleep periods 1-4, respectively, 14.3 ± 5.2, 
24.8 ± 6.9, 21.5 ± 12.6, 19.5 ± 3.9). Data from all subjects 
have been transformed to this common mean time scale 
before averaging. Average sleep stages were determined 
by associating values of  -1, 0, 1, 2, 3 and 4 to stages of 
wakefulness, REM sleep, sleep stage 1, 2, 3 and 4. The 
variations during REM sleep indicate transient shifts into 
NREM sleep stages or wakefulness. n = 6.  

were in all cases transitory, and the original potential level 
was recovered immediately after the end of saccadic eye 
movements, which made an exclusion of these periods 
unnecessary. Care was taken to maintain constant skin 
temperature at all electrode locations. This was most 
successful as can be seen by the temperature courses in 
Figs 6 and 7. In addition, the lack of correlation shows that 
influences of temperature on the course of the DC potential  

Fig. 5  A schematic time course of DC potential changes 
during the NREM-REM sleep cycle. Four phases can be 
discriminated: (1) a steep NREM transition negative shift, 
(2) a gradual NREM positive slope, (3) a steep  REM
transition positive shift, and (4) a REM negative slope.  
 
 
were negligible. In this context, results of the previous 
study (Marshall et al. 1996b) also revealed that any strong 
influences of gross changes in blood gas tension (as 
derived from end expiratory CO

2

 concentration) on sleep

associated DC potential shifts could be excluded. 
Although, under extreme conditions of hypo  and 
hypercapnia, DC potential shifts can be readily discerned 
(Caspers  et al. 1987;  Rockstroh 1990). Thus, from the 
above mentioned, in conjunction with other studies, it can 
be assumed that the principle source of the DC potential 
shifts reported here derives from neuronal activity, with a 
possible contribution of glia cells (e.g.  Caspers  et al
1984; Birbaumer et al. 1990).  

The significance of the present data derives from the 

fact that this study monitored DC potential changes during 
the entire period of nocturnal sleep also  including the 
epochs of wakefulness preceding and succeeding sleep. 
Comparing the DC potential during sleep and adjacent 
periods of wakefulness indicated higher negative DC 
voltage during sleep than morning wakefulness. 
Considering the susceptibility of DC potential recordings 
to artefacts inducing linear trends, the primary analyses 
here, based on data after removal of such trends, 
confirmed higher DC negativity during sleep as compared 
to morning wakefulness and also as compared to evening 
wakefulness, over frontocortical areas. This finding 
contrasted our hypothesis, but is consistent with the report 
of a sleep related negative DC potential shift in humans 
by  (Hoffmann  et al. 1988) and coincides with the 
increment in negativity found at sleep onset in previous 
studies (Marshall et al. 1994, Marshall et al. 1996b).  

More prominent than the sleep associated increase in 

average negativity of the DC potential level was the 
dynamic regulation of DC potential changes, i.e. the 
recurrent pattern of negative and positive DC potential 
shifts in the course of nocturnal sleep. The underlying 
regulatory mechanism of these shifts is at present 
unclear. 

The slower NREM positive and REM negative slopes 

may reflect two sustained states resulting from ongoing 
activity in the respective sleep stages rather than activity 
of the putative NREM and REM sleep generating 
processes per se. In the case of the REM negative slope, 
the gradual negative slope in the DC potential may  

 

L. Marshall et al.    DC potential shifts during nocturnal sleep               1175 

0

20 40 60 80 100 120 140 160

°C

30

32

34

A

B

D

right mastoid

left mastoid
central

m2

0

20 40 60 80 100 120 140 160

mV*mV/Hz

0.0

0.5

1.0

1.5

0

20 40 60 80 100 120 140 160

mV

-0.2

-0.1

0.0

0.1

0.2

0

20 40 60 80 100 120 140 160

4

3

2

1

REM

W

NREM

REM

NREM

NREM-transition-negative shift

NREM-positive slope

REM-transition-positive shift

REM-negative slope

min

central 

frontal 

C

0

20 40 60 80 100 120 140 160

mV

2

/Hz

0

1

2

3

4

5

0

20 40 60 80 100 120 140 160

mV

-0.2

0.0

0.2

0.4

0

20 40 60 80 100 120 140 160

4

3

2

1

REM

W

0

20 40 60 80 100 120 140 160

30

32

34

C

D

°C

min

B

A

NREM

REM

NREM

NREM-transition-negative shift

NREM-positive slope

REM-transition-positive shift

REM-negative slope

right mastoid

left mastoid

central

central 

frontal 

Fig. 6  Averaged time courses of sleep stage (A), DC 
potential shifts (after linear trend elimination, B), slow wave 
activity (C), and scalp temperature (D) across the second 
to fourth NREM-REM sleep cycle. Thick lines represent 
averages of F3 and F4 (frontal), and thin lines represent 
averages of C3 and C4 (central). The durations of the 
NREM and REM sleep periods correspond to the average 
duration of each of these intervals, which was (mean ± 
SEM) 81.8 ± 5.2 min and 23.7 ± 4.0 min, respectively. Data 
of all subjects have been transformed to this common 
mean time scale before averaging. (For clarity, the time 
course of the parameters has been extended to include the 
NREM period following the REM sleep period.) The four 
phases characterizing the DC potential changes across the 
NREM-REM cycle are indicated by vertical lines (refer to 
Fig. 5). n = 12.  

 

indicate the cumulative effect of disinhibited intralaminar or 
ventromedial thalamocortical input. These nonspecific 
nuclei project to apical dendrites of cortical layer I, the 
depolarization of which is considered a main source of 
cortical surface negative potentials during wakefulness 
(Birbaumer  et al. 1990). Cholinergic projections to the 
neocortex are also believed to be essential for the 
generation of negative DC potential shifts in wakefulness 
(Birbaumer  et al. 1990). Since in REM sleep cortically 
projecting, presumably cholinergic neurons become 
(tonically) more activated than in NREM sleep (Nuñez 
1996), this system may also contribute to the REM
negative DC potential slope. Conceivable also is that a 
cholinergic enhancement of cortical excitability facilitates 
the formation of fast alternating oscillations (Llinás & Ribary 
1993; Steriade et al. 1996). Conversely, in the case of the 
NREM positive slope, the gradual positive DC potential 
change may reflect a disfacilitation of cortical excitability 
possibly linked to an increasing inhibition of cholinergic 
input to the cortex.  

Fig. 7  Averaged time courses of sleep stage (A), DC 
potential shifts (after linear trend elimination, B), slow 
wave activity (C), and scalp temperature (D) across the 
first NREM-REM sleep cycle. Thick lines represent 
averages of F3 and F4 (frontal), and thin lines represent 
averages of C3 and C4 (central). The durations of the 
NREM and REM sleep periods correspond to the average 
duration of each of  these intervals, which was (mean ± 
SEM) 80.2 ± 7.6 min and 13.8 ± 4.5 min, respectively. 
(For clarity the time course of the parameters has been 
extended to include the NREM period following the REM 
sleep period.) Data of all subjects have been transformed 
to this common mean time scale before averaging. The 
four phases characterizing the DC potential changes 
across the NREM-REM cycle are indicated by vertical 
lines (refer to Fig. 5). n = 7. 

 
Although the REM negative slope shifted toward 

negative values, the NREM positive slope on  average 
maintained a more negative level than the REM negative 
slope. At first glance, this difference appears difficult to fit 
into the concept of REM sleep representing a state of 
higher brain activation than NREM sleep and slow wave 
sleep. However, the association of the DC potential with 
increased cortical excitability due to widespread 
depolarization of apical dendrites has been derived from 
studies in awake subjects, and different mechanisms 
would be relevant for DC potential changes in sleep. 

A different mechanism of regulation of sleep related DC 

potential shifts than those of scalp recorded DC potential 
shifts in wakefulness, e.g. during preparatory states, is 
also suggested by comparing magnitudes of these shifts, 
which in the latter are one to several hundred microvolts 
smaller (Haider et al. 1981; Pleydell Pearce et al. 1995). 
Thus, rather than differences in apical depolarization, the 
transitional shifts which draw the average potential level 
during NREM sleep towards negative values and during 
REM sleep towards positive values, could express the