10.

2478/msr-2014-0032
MEASUREMENT SCIENCE REVIEW, Volume 14, No. 4, 2014

Spectral EEG Features of a Short Psycho-physiological Relaxation

Michal Teplan1 , Anna Krakovská1 and Marián Špajdel2,3

1 Institute of Measurement Science, Slovak Academy of Sciences, Bratislava, Slovakia, [email protected]
2 Institute of Normal and Pathological Physiology, Slovak Academy of Sciences, Bratislava, Slovakia
3 Faculty of Philosophy and Arts, University of Trnava, Slovakia

Short-lasting psycho-physiological relaxation was investigated through an analysis of its bipolar electroencephalographic (EEG)
characteristics. In 8 subjects, 6-channel EEG data of 3-minute duration were recorded during 88 relaxation sessions. Time course of
spectral EEG features was examined. Alpha powers were decreasing during resting conditions of 3-minute sessions in lying position
with eyes closed. This was followed by a decrease of total power in centro-parietal cortex regions and an increase of beta power in
fronto-central areas. Represented by EEG coherences the interhemispheric communication between the parieto-occipital regions
was enhanced within a frequency range of 2-10 Hz. In order to discern between higher and lower levels of relaxation distinguished
according to self-rated satisfaction, EEG features were assessed and discriminating parameters were identified. Successful relax-
ation was determined mainly by the presence of decreased delta-1 power across the cortex. Potential applications for these findings
include the clinical, pharmacological, and stress management fields.
Keywords: EEG, relaxation, biosignal analysis, coherences.

1. I NTRODUCTION meditation and during or after such a state. However, here

HE RELAXATION response is an integrated body reaction we focused on a gradual change of the physiological state.
T that reflects the voluntary resting state of both the body
and the mind. The characterization of the changes in brain ac-
Various types of meditation are known to induce a relaxed
state [5, 6, 1]. Usually, with increased level of meditation,
tivity during the resting state is relevant to many fields, includ- a deeper state of relaxation can be established. This finding
ing stress reduction, sleep deprivation, and the testing of phar- is supported in parallel by subjective observations and phys-
macological substances related to drugs with hypnotic and iological parameters (e.g., respiratory rate, skin conductance,
sedative effects. Relaxation response influences mental and and production of plasma lactate) [1].
physical health and has an important role in moderating ten- However, the evidence presented in the literature is incon-
sion and psychosocial stress. In a broader sense, stress is ac- sistent regarding the characterization of relaxation using spe-
knowledged as one of the major problems of modern society. cific EEG features. The increased power of the alpha and theta
Regular relaxation practice can affect the various physiolog- frequencies [5, 7, 8, 3, 9] and interhemispheric synchroniza-
ical and psychological parameters involved in aging, diges- tion, especially frontal alpha coherence [1], are usually con-
tion, general well-being, and psycho-somatic disease. Conse- sidered to be neurophysiological indicators of a state of sen-
quently, there is a growing need to monitor the physiological sorimotor and mental rest.
processes associated with relaxation and stress response. In our previous study, we observed an increase in the al-
The term "relaxation" conveys positive and beneficial pha and theta powers and an enhanced frontal alpha coherence
phenomena. However, it is difficult to define psycho- [10] after two months of audio-visual stimulation training (25
physiological relaxation in terms of specific physiological pa- repetitions of 30-minute intervals of relaxation) [11].
rameters. According to Travis [1], five categories of physio- The aim of the present study was to directly examine the
logical variables have the potential to discern between the lev- spectral EEG characteristics during a 3-minute interval while
els of relaxation, including breath and heart rates, as well as the subjects were in a resting state. In particular, two sequen-
skin conductance. Similar parameters for the characterization tial tasks were addressed: 1) the detection of changes in spec-
of stress levels were reported by Vavrinský [2]. Physiologi- tral EEG measures during 3-minute general sensorimotor re-
cal indicators of the relaxation response introduced in Foster laxation and 2) the selection of EEG features that are able
[3] include decreased oxygen consumption, respiration, and to distinguish between self-rated successful relaxation and re-
heart rate, in addition to increased alpha brain wave produc- laxation which was subjectively perceived as less efficient.
tion. According to Ossebaard [4], relaxation is comprised of
biological, psychological, and social components.
2. M ATERIAL AND METHODS
Another useful source of information on relaxation is re-
search on meditation. In fact there are no issues focused di- 2.1. EEG recording
rectly on EEG dynamics during relaxation induction in the lit- Eight right-handed, healthy subjects (3 females and 5 males,
erature. Differences between two distinct states or conditions mean age 27 years, SD 7 years) volunteered for the experi-
are usually investigated, e.g., before the state of relaxation or ment. The subjects were instructed to lie down comfortably

237
 MEASUREMENT SCIENCE REVIEW, Volume 14, No. 4, 2014

overlap. All of the calculations were carried out in the Mat-

lab environment (by Mathworks, version R2008b) with the
FpZ use of the pwelch and cpsd functions. The powers were an-
F3 F4
alyzed in 6 cortical locations and were obtained by the off-
C3 CZ C4 line data transformation into bipolar mode as described above.
The coherences were investigated between 11 different pairs
P3 P4
(F3C3-F4C4, etc.) of the 6 scalp regions. Combinations with
O1 O2 common electrodes (e.g., F3C3-C3P3) were not considered in
order to avoid spurious results caused by contribution of the
common electrode into both signals.
Each subject participated in 24 recording sessions and at-
tended only one session per working day. The data were col-
Fig. 1: The EEG montage used in this study is shown. The active
lected over a 2-month period. Prior to participation, every
electrodes were placed at the F3, F4, C3, C4, P3, P4, O1, and O2
locations with the reference and ground electrodes at the Cz and Fpz subject completed an entrance form assessing general health
points, respectively. status, medication, involvement, motivations and expectations
of the experiment. One of the eight participants reported prior
on their backs, keeping their eyes closed and relax both phys- experience in meditation, yoga or other relaxation techniques.
ically and mentally. No further instructions or technique were Before each session, participants filled out a one-page struc-
provided regarding how to establish a relaxing state. The ex- tured form where their actual overall status was reported in a
periment was conducted in a darkened, electrically shielded form of health, emotional and mental components. Also their
room. The subjects placed in these conditions were expected medication and caffeine or nicotine intake was questioned.
to realize a state of sensory, motor, as well as mental relax- The subjects were asked to rate the progress in relaxation at-
ation. tained during the 3-minute resting period after each recording
The EEG data for a 3-minute interval were recorded from session. A seven-point bipolar Likert scale [12] was used for
eight scalp positions according to the International 10-20 sys- assessment. The relaxation extremes rated by the scale ranged
tem: F3, F4, C3, C4, P3, P4, O1, and O2 (Figure 1). The ref- from "tense or aroused" ( -3) to "relaxed" (+3).
erence electrode was located at Cz and the ground electrode at The time course of the various EEG measures (powers and
Fpz. The Ag/AgCl electrodes were placed on a standard EEG coherences) during each 3-minute interval was determined
cap (Electro Cap Inc.). Impedances were kept below 5 kΩ and based on the spectra, which were computed in 2-second sub
balanced within 1 kΩ. The parameters for our prototype EEG windows. The resulting curves were fitted using linear re-
recording unit were as follows: amplifying gain: 402; sam- gression. The magnitude of the time trends was evaluated by
pling frequency: 500 Hz; A/D converter resolution: 14 bits; F-test values that were obtained from testing whether the re-
input resolution: 0.46 µ V; noise: max 4.1 µ V pp. (0.07 to gression slope differs from the zero one.
234 Hz); low pass filter: 234 Hz (-3 dB); and high pass filter:
0.07 Hz (-3 dB).
3. R ESULTS
From the 8 signals recorded from active electrodes, six
bipolar signals (F3-C3, F4-C4, C3-P3, C4-P4, P3-O1, and 3.1. General relaxation
P4-O2) were derived off-line in order to avoid the undesirable In this part we examined all EEG data that were measured,
effects of the common reference electrode. All of the signals regardless of subjects’ self-rated satisfaction. 11 artefact-free
were visually inspected; the sequences containing artefacts or recordings from each participant were selected. For each mea-
sleep phenomena were excluded with support of information sure and cortex location, we had a total of 88 time curves from
from forms filled in by participants. Preprocessing was com- the 8 subjects. For most of the EEG measures, a linear time
pleted by the application of a digital high pass FIR filter with a course was observed during the 3-minute recording period. In
cut-off frequency of 0.5 Hz and a 3000-point Blackman win- order to identify the prominent EEG features that character-
dow. ize the relaxation process, we focused on the strongest time
trends. All time trends were quantified by F-test values de-
rived from evaluation of differences of the regression slopes
2.2. Measures from the zero slope. The F-values were modified with ad-
We analyzed the total power, relative power, and magnitude dition of sign designating decreasing (minus) or increasing
squared coherence in nine frequency bands: delta-1 (0.5-2 (plus) trends. Modified F values were not distributed nor-
Hz), delta-2 (2-4 Hz), theta-1 (4-6 Hz), theta-2 (6-8 Hz), mally. Therefore, we used nonparametric testing to identify
alpha-1 (8-10 Hz), alpha-2 (10-12 Hz), beta-1 (12-16 Hz), changes in the EEG powers during resting conditions. A sign
beta-2 (16-30 Hz), and gamma (30-45 Hz). The total power test was applied to test whether medians differed from zero.
ranged from 0.5 to 45 Hz. The time courses are illustrated as average time curves
The frequency analysis was calculated using a fast Fourier (n=88) to highlight the measures of the most significant
transform within a 2-second window with 0.75 seconds of changes within the sessions. A decrease in total EEG power

238
 MEASUREMENT SCIENCE REVIEW, Volume 14, No. 4, 2014

3−minute rest: 3−minute relaxation:

relative alpha−1 power
total power
%
2
µV 19 P4−O2
C3−P3
55 18
17
50 16
15
45 14
13
40
50 100 150 50 100 150
time [s]
time [s]
3−minute relaxation:
relative band powers
Fig. 2: The total EEG power changes in the C3-P3 location during %
general relaxation are shown for the 8 subjects. The grand average P4−O2
20
time course, which was calculated by averaging the 88 curves from
each of the recording sessions, is presented.
15
was observed in the centro-parietal location C3-P3 (Figure 2)
and C4-P4 (p-values < 0.02 and 0.05, respectively). 10
The strongest decrease was related to the alpha-1 wave con-
tribution. With the exception of the fronto-central areas, the 5
relative alpha-1 power dropped in all of the locations: C3-P3,
C4-P4, P3-O1, and P4-O2 (all p-values < 10−5 , Figure 3). 0
The most significant increase was found for the beta-1 rel- δ1 δ2 θ1 θ2 α1 α2 β1 β2 γ
ative power in the fronto-central F3-C3 (Figure 4) and F4-
C4 areas (p-values < 10−5 and 10−3 , respectively). More-
over, the higher frequency bands, beta-2 and gamma, also in- Fig. 3: Above: The time course for the grand average alpha-1 rel-
ative band power in the P4-O2 location during general relaxation.
creased their expression in the F3-C3 area (p-values < 0.008
Bellow: The grand average relative band powers. The data for the
and 0.004, respectively; Figure 4). initial (white) and closing periods (gray and black) are presented for
The most prominent EEG effects were realized within the the P4-O2 cortex location. Significant changes in the alpha-1 band
alpha-1 waves in the P4-O2 area (Figure 3); the average rela- are highlighted in black.
tive alpha-1 power was diminished by 30 % of its initial value.
This measure decreased in 76% of all the single measurement two sets according to the subjects’ self-ratings of the relax-
sessions. The decrease of the relative alpha-1 power often led ation progress. Out of 11 recordings from each subject the
to positive relative increments for all the other bands. This 4 records with the highest ratings and the 4 records with the
fact is evident in the bar graphs shown in Figures 3 and 4, lowest ratings of attained relaxation level were chosen. The
which compare the spectral band characteristics in a form of subjects differed in their strategies for the relaxation level rat-
grand averages from the initial and the final 20-second inter- ings. Some of the subjects reported extreme values (-3 or 3),
vals of the 3-minute recording period. while some reported only values closer to zero (-1,0,1). The
Significant changes were also identified in the coherence rating distribution was asymmetrical with the positive ratings
data. The interhemispheric communication between the occurring more frequently. The mean rating values were -0.06
parieto-occipital centres was enhanced. This conclusion was for less successful relaxation and 1.88 for more successful re-
based on the increase in the P3O1-P4O2 coherence in the four laxation.
neighbouring power bands, which include the delta-2, theta-1
(Figure 5), theta-2, and alpha-1 (all p-values < 0.008). For all of the participants, the linear regression was calcu-
lated separately for each of the four recordings given his or
her extreme low or high relaxation self-rating. We focused
3.2. Successful relaxation on a comparison of the two groups according to the following
Our second goal was to determine which of the EEG measures three parameters derived from the time course of the single
are associated with self-rated successful relaxation. There- EEG spectral characteristics. Two of them were the initial and
fore, the recordings from the previous section used for spec- final values obtained from linear regression of the time course
ifying the state of general relaxation were divided here into accounted for the first and last 6 seconds of the recordings, re-

239
 MEASUREMENT SCIENCE REVIEW, Volume 14, No. 4, 2014

3−minute relaxation: 3−minute rest:

relative beta−1 power theta−1 coherence
% 0.42
7 F3−C3 P3−O1/P4−O2

0.4
6.5

6 0.38

5.5 0.36

5 0.34
50 100 150 50 100 150
time [s] time [s]

3−minute relaxation:
relative band powers
Fig. 5: The time course of grand average for theta-1 coherence be-
% tween the P3-O1 and P4-O2 locations during general relaxation is
20 F3−C3 shown. The mean time course was calculated by averaging the 88
curves from each of the recording sessions.
15
in the beta-1 band (between the F4C4 and C3P3 locations,
10
p = 0.017). Different relaxation states displayed also differ-
ent regression slopes according to the modified F-values. For
more successful relaxation, rise of the alpha-2 coherence be-
5
tween the F3C3 and P3O1 locations (p < 0.002, Figure 6)
and decrease of the coherence in the beta-1 range between the
0
δ1 δ2 θ1 θ2 α1 α2 β1 β2 γ C3P3 and C4P4 locations (p < 0.05) was observed.

4. D ISCUSSION
Fig. 4: Above: The time course for the grand average beta-1 rela- A decrease in the total power across the entire cortex dur-
tive band power in the F3-C3 location during general relaxation is ing the relaxation condition implies that, overall, the subjects’
shown. Bellow: The grand average relative band powers are shown. brain activity gradually diminished along the relaxation pro-
The initial (white) and closing periods (gray and black) in the F3-C3 cess. We anticipated an increase in the alpha powers [3, 7, 8];
cortex location are depicted. The bands with significant changes are but surprisingly, we discovered a pronounced and significant
highlighted in black.
decrease in alpha-1 power, which was a definite leading factor
spectively. The slopes were represented by the sign modified in the observed decline in overall performance (Figure 3).
F-values introduced in the previous section. In order to find In the literature, it is a commonly accepted fact that the
the distinct features of the two relaxation categories, a non- increase in the alpha waves is associated with rest and relax-
parametric Kruskall-Wallis test was used for the 2 groups (n ation [3]. However, this phenomenon may be partly related to
= 32; 4 sessions for each of the 8 subjects). The significance closing the eyes. Closing the eyes usually initiates rest states,
threshold was p < 0.05. which can occur for short periods of time, even when the sub-
Deeper relaxation was associated with lower final values ject is sitting or standing. In our study, the relative power of
of the slowest spectral components. Lower contribution of the alpha-2 band was increased for all of the cortex regions,
delta-1 waves was present in the following derivations: F4- but only during the first 30 seconds. This increase was fol-
C4, C3-P3, and P3-O1 (all p-values < 0.03, Figures 6 and 7). lowed by a strong decline in the power of the alpha-1 waves
In the three remaining cortex locations, the delta-1 shift for for the remainder of the 3-minute relaxation period.
successful relaxation was lower as well, however, not statis- However, a decrease in the relative alpha power did not
tically significant. Already at the beginning of the relaxation guarantee that successful relaxation was achieved. Self-rated
sessions, the beta-2 power in the P3-O1 location was higher satisfaction with a relaxation state correlated with the de-
for the more successful relaxation sessions than for the less creased expression of the delta-1 relative power.
successful relaxation sessions (p < 0.04). Indicators of enhanced relaxation in a form of lower contri-
For certain coherences, lower final values were obtained in- bution of the slowest waves remove any speculations about the
terhemispherically. Specifically, decrement was found for the correlation of the observed relaxation states with periods with
better relaxation in case of the alpha-1 coherence (between eventual occurrence of sleep onset. Sleeping states are char-
the C3P3 and P4O2 locations, p < 0.04) and the coherence acterized instead by the increased occurrence of delta waves.

240
 MEASUREMENT SCIENCE REVIEW, Volume 14, No. 4, 2014

Less and more successful relaxation:

Less and more successful relaxation: relative band powers
relative delta−1 power %
25
% P3−O1
C3−P3
23 20

15

19 10

5
15
0
δ1 δ2 θ1 θ2 α1 α2 β1 β2 γ
50 100 150
time [s]

Fig. 7: A comparison of the grand average relative band powers

Less and more successful relaxation:
alpha−2 coherence at the end of the 3-minute relaxation period in the P3-O1 location
are shown. Less successful relaxation (white) and more successful
0.34 F3−C3/P3−O1 relaxation (grey and black). Black represents statistically significant
difference.

cordance with the data from questionnaires in which the par-

0.3 ticipants occasionally reported their thought processes. Re-
garding this fact, our aim was not to study relaxation under
restricted conditions; for example, meditation studies focus
on highly trained subjects. In contrast, our goal was to in-
0.26 vestigate the relaxation processes in normal adult population.
For most individuals, pleasant and beneficial relaxation can
50 100 150
time [s] involve episodes with relaxed thought processes. Hence, we
instructed the participants to relax generally, which includes
physical, mental, and emotional aspects, and we did not pro-
vide any special instructions or methods specifying how to
Fig. 6: The difference between less (gray) and more (black) success-
ful relaxation is illustrated. The grand average of time course was relax.
calculated by averaging all of the curves for each of the recording Upon completion of this experiment, half of the partici-
sessions. The difference in the final values for relative delta-1 power pants reported their capacity for relaxation to be improved
in C3-P3 (above) and in the slopes for alpha-2 coherence between due to repetitive participation in the relaxation experiment
F3-C3 and P3-O1 (bellow) are shown. sessions.
During the general relaxation process we observed an in- We have introduced a new approach to the investigation of
crease in the coherences between the hemispheres at the relaxation through time course of EEG features. Furthermore,
parieto-occipital sites in the 2-10 Hz frequency range and a we have confronted general assertions on relaxation with the
decrease in some of the beta-1 coherences interhemispheri- computed spectral EEG characteristics. In this sense, our re-
cally. However, it is not possible to make a clear conclusion sults contribute to the characterization of relaxation processes
regarding the ability of coherence to discern between the more with the use of spectral EEG features and might be benefi-
and less successful relaxation states. cial for further research in areas like neurofeedback and rest-
This conclusion is in accordance with the findings of Gay- ing states analysis. In the case of neurofeedback, the features
lord [13], who reported that the coherence was higher in the found to be characteristic for the state of relaxation might be
frontal and central areas only during transcendental medita- used in biofeedback training protocol in order to improve re-
tion practice relative to the closed eyes condition, but not dur- laxation induction.
ing progressive muscle relaxation. Moreover, we have shown that with the use of EEG data it
Several of the subjects occasionally reported some form of is possible to discriminate between two relaxation categories
thinking (often related to their work or everyday problems). formed according to the quality of relaxation. We have found
It can be assumed that internal thought would occur regularly that the focus should be directed to the slowest brain waves. It
for most of the participants, regardless of their awareness of is important to note that our results may be specific to the con-
it, as the subjects had no special training in controlling their ditions of our study; therefore, different EEG outcomes may
mental activity during the relaxation. This finding is in ac- result from a different set of relaxation conditions. Another

241
 MEASUREMENT SCIENCE REVIEW, Volume 14, No. 4, 2014

factor influencing outcomes of the study might be a choice of [5] Banquet, J. (1973). Spectral analysis of the EEG in
unipolar or bipolar EEG due to data processing. meditation. Electroencephalography and Clinical Neu-
For future studies we suggest simultaneous monitoring of rophysiology 35, 143–151.
additional physiological variables (e.g., cardiovascular, res- [6] Aftanas, L., Golocheikine, S. (2001). Human anterior
piratory, muscular, skin conductance, and blood parameters and frontal midline theta and lower alpha reflect emo-
responsive to the level of stress) so that the EEG parameters tionally positive state and internalized attention: high-
can be associated with complex physiological responses more resolution EEG investigation of meditation. Neuro-
directly. Also a more prolonged relaxation period could be in- science Letters 310, 57–60.
vestigated, and subjects’ mental processes could be monitored [7] Lindsley, D. (1952). Psychological phenomena and
by more specific questionnaires. the electroencephalogram. Electroencephalography and
Clinical Neurophysiology 4, 443–456.
5. ACKNOWLEDGEMENT [8] Brown, B. (1970). Recognition of aspects of conscious-
The research was supported by the grant VEGA 2/0043/13. ness through association with EEG alpha activity repre-
sented by a light signal. Psychophysiology 6, 442–452.
R EFERENCES [9] Jacobs, G., Friedman, R. (2004). EEG Spectral Analy-
[1] Travis, F. (2001). Autonomic and EEG patterns distin- sis of Relaxation Techniques. Applied Psychophysiology
guish transcending from other experiences during Tran- and Biofeedback 29, 245–254.
scendental meditation. International Journal of Psy- [10] Teplan, M., et al. (2006). EEG responses to long-term
chophysiology 42, 1–9. audio-visual stimulation. International Journal of Psy-
[2] Vavrinský, E. (2005). Thin-film microelectrodes for chophysiology 59, 81–90.
electro-chemical conductive sensors applied in biomed- [11] Teplan, M., et al. (2011). Direct and transient effects
ical monitoring of stress. Ph.D. thesis, Slovak Technical of audio-visual stimulation on EEG. Computer Methods
University in Bratislava, Faculty of Electrical Engineer- and Programs in Biomedicine 102, 17–24.
ing and Informatics, Department of microelectronics. [12] Likert, R. (1932). A Technique for the Measurement of
[3] Foster, D. (1990). EEG and subjective correlates Attitudes. Archives of Psychology 140, 1–55.
of alpha-frequency binaural-beat stimulation combined [13] Gaylord, C., et al. (1989). The effects of the transcen-
with alpha biofeedback. Memphis State University. dental meditation technique and progressive muscle re-
[4] Ossebaard, H. (2000). Stress reduction by technology? laxation on EEG coherence, stress reactivity, and mental
An experimental study into the effects of brainmachines health in black adults. International Journal of Neuro-
on burnout and state anxiety. Applied Psychophysiology science 46, 77–86.
and Biofeedback 25, 93–101.
Received September 14, 2013.
Accepted July 31, 2014.

242