You are on page 1of 4

Neuroscience Letters 267 (1999) 58

EEG patterns in theta and gamma frequency range and their


probable relation to human voluntary movement organization
David Popivanov a,*, Antoaneta Mineva a, Ivan Krekule b
a
Institute of Physiology, Bulgarian Academy of Sciences, 1113 Soa, Bulgaria
Institute of Physiology, Acamedy of Sciences of The Czech Republic, 142-02 Prague, Czech Republic

Received 3 February 1999; received in revised form 12 March 1999; accepted 12 March 1999

Abstract
In experiments with EEG accompanying continuous slow goal-directed voluntary movements we found abrupt shortterm transients (STs) of the coefcients of EEG time-varying autoregressive (TVAR) model. The onset of STs indicated (i)
a positive EEG wave related to an increase of 37 Hz oscillations in time period before the movement start, (ii) synchronization of 3540 Hz prior to movement start and during the movement when the target is nearly reached. Both these
phenomena are expressed predominantly over supplementary motor area, premotor and parietal cortices. These
patterns were detected after averaging of EEG segments synchronized to the abrupt changes of the TVAR coefcients
computed in the time course of EEG single records. The results are discussed regarding the cognitive aspect of organization of goal-directed movements. q 1999 Published by Elsevier Science Ltd. All rights reserved.
Keywords: EEG and goal-directed movements; Theta and gamma oscillations; Positive wave; Autoregressive time-varying model;
Voluntary movement organization

The voluntary movement performance is preceded by


specic EEG activity patterns, which are endogenous and
considered to reect cognitive aspects of the movement
preparation. The most typical feature of the preparatory
process (negative trend-like shift of EEG preceding the
movement onset) is referred to as Bereitschaftspotential
[2]. Another typical phenomenon prior to voluntary movement is a decrease of both alpha and beta frequency components over the sensorimotor areas, accompanied by
enhanced gamma oscillations [1,1017]. As the onsets of
the trend and these frequency changes do not always coincide in time, it is hard to distinguish precisely those EEG
dynamic changes that reect the onset of movement
preparatory period and its consecutive phases.
Autoregressive (AR) models have been successfully used
to describe linear dynamics of EEG activity [5,8,11]. In
previous studies [11,14], using time-varying autoregressive
(TVAR) models of EEG accompanying voluntary movements, we found abrupt short transients of the model coefcients (STs) during the movement preparatory period,
related to time-frequency dynamic changes in alpha, beta
* Corresponding author. Tel.: 1359-2-705-259; fax: 1359-2719-109.
E-mail address: dapo@iph.bio.bas.bg (D. Popivanov)

and gamma bands. The results suggest that these STs (indicating temporal nonstationarity of the EEG segments) probably reect boundaries separating successive phases of the
movement preparation.
The objective of this study is to further attempt to analyze
the EEG changes related to the voluntary movement organization in experiments with goal-directed movements. The
main questions addressed in the study were, (i) whether the
STs indicate systematically typical changes in EEG
frequency components, (ii) whether the averaged EEG
segments synchronized to STs, will reveal EEG patterns
time-locked to the phases of the movement preparation.
Six right-handed healthy volunteers (aged 2030 years,
four males, two females) took part in the experiment
designed for a study of slow goal-directed movement
under conditions of a free choice of target and instance of
action. The subjects faced a vertical screen that contained
three targets and positioned the right-hand ngers (right
elbow supported) on a manipulator with three degrees of
freedom, the rotation of which directed a light beam to the
screen. The subject was required to choose one of the
targets, to direct the beam to it and to press a micro-switch
when reaching the target. Precision rather than speed of
reaching was demanded. The trials were repeated at irregular intervals not shorter than 20 s. The EEG was recorded by

0304-3940/99/$ - see front matter q 1999 Published by Elsevier Science Ltd. All rights reserved.
PII: S03 04 - 394 0( 9 9) 00 27 1- 2

D. Popivanov et al. / Neuroscience Letters 267 (1999) 58

Fig. 1. Top: single EEG record (Fz) and markers (three vertical
bars) of the movement phases (start at 0 s, switch pressing,
returning to initial position). Middle: traces of the TVAR coefcients and markers (two vertical bars) of the selected short transients (before the start (ST1)and before the target reaching
(ST2)). Bottom: momentary EEG spectra computed by TVAR
coefcients before (dashed lines) and immediately after (solid
lines) ST1 (left) and ST2 (right).

Ag/AgCl electrodes from F3, Fz, F4, C3 0 , Cz, C4 0 , P3, Pz,


P4 (10/20system, C3 0 , C4 0 positioned 1 cm anterior with
respect to the standard C3 and C4 location) with linked
earlobes as reference. The EOG was recorded bipolarly to

Fig. 2. Top to bottom: averaged EEG signal (F3) synchronized to


ST1 (one subject). Filtered signal: 37 and 3540 Hz. A positive
wave (1st trace) and time-locked synchronization of theta and
gamma bands (2nd and 3rd traces) are seen at ST1 marked by
vertical lines at 0 s.

detect both vertical and horizontal eye movements. The


EEG and EOG signals were amplied using Nihon Kohden
EEG - 4314F (upper cut-off frequency (23 dB) of 120 Hz;
time constant of 5 s) and digitized together with markers of
the movements performance (10 bit A/D converter, 256
samples/s) [6]. The data records were synchronized to the
marker of the movement start (17 s epochs, 10 s prior to the
movement start). Only artifact-free EEG records were
processed.
The following procedure was applied to data analysis, (i)
the single EEG record was normalized to zero mean and unit
variance. The appropriate AR model order was determined
using Burg algorithm and Rissanen's minimum description
length criterion [9,11] from an EEG segment of 2 s at the
beginning of the record with no obvious nonstationarities
long before the assumed movement preparatory period. For
the xed AR model (it was in the range of 612), the
variance of the residuals VE was computed. (ii) An algorithm
of Kalman lter type was used to compute the coefcients of
the TVAR model in steps along the EEG record. The algorithm updates recursively several models and selects that
one with highest posterior probability [9]. The variance of
innovations in the model R2 and the probability q that the
model undergoes abrupt changes at any given time were
determined empirically: R2 0:63VE; ; q 0:01. (iii) The
TVAR coefcients were plotted over time, revealing abrupt
STs of all coefcients in some instants, disturbing almost
constant levels of the coefcient traces. Short-term transients with the largest amplitude above a preselected threshold and duration of about 100 ms were selected and their
onsets were stored. (iv) The EEG records were synchronized
to the STs and averaged (3550 selected EEG records per
subject). If more than one ST existed in a given segment the
rst one was taken supposing to indicate the onset of a given
phase of the movement organization. Thus, for each site,
two averaged EEG segments of 4 s were obtained, one
synchronized to ST1 (before the movement onset) and the
second to ST2 (before the switch pressing). Fig. 1 illustrates a single EEG record, markers of the movement, the
behavior of TVAR coefcients and momentary EEG spectra
around selected ST1 and ST2. Peaks in theta range after ST1
(left bottom) and in gamma range after ST2 (right bottom)
are well pronounced (thick lines, see below).
To examine the frequency changes a narrow pass-band
ltering without phase distortion was applied to averaged
segments. The frequency bands were individually specied,
but most typical were the ranges of 37 Hz (theta), 812 Hz
(alpha), 1730 Hz (beta)and 3540 Hz (gamma).
After averaging of the EEG synchronized to the ST1
onsets a positive wave (PW) was observed around the
ST1. The PW was with an amplitude in a range of 10 mV
and a duration of about 150 ms and was followed by a
negative shift. A decrease of the EEG variance was
observed prior to the PW as well. The narrow pass-band
ltering in the vicinity of PW revealed an increase of
theta activity (37 Hz) following the PW, being most promi-

D. Popivanov et al. / Neuroscience Letters 267 (1999) 58


Table 1
Range of signicant frequency synchronization (Wilcoxon rank
test, P , 0:05) for all subjects
Electrode position

After ST1 (Hz)

F3
Fz
F4
C3 0
Cz
C4 0
P3
Pz
P4

37
37
37
37
37

After ST2 (Hz)

3540
3540
3540

37

3540
3540
3540

37
37
37
37

3540
3540
3540
3540
3540
3540

nent at contralateral sites (F3, C3 0 , P3), and at Fz and F4


sites in some cases. The PW was often accompanied by
bursts of gamma (3540 Hz) activity expressed mainly
over frontal and parietal areas (Fig. 2 and Table 1).
After EEG averaging with respect to ST2 less consistent
PW was also seen (mainly at C3 0 , C4 0 and over the parietal
cortex) and its frequency content was also in the theta
range (Table 1). The trend became positive and in some
subjects a weak negativity followed the ST2. The most
important nding was the abrupt and enhanced gamma
(3540 Hz) bursts starting immediately after ST2 onset
and preceded by a clear suppression of this activity for
intervals longer than 500 ms (Fig.1, bottom). The enhancement of gamma activity was observed (and started almost
in parallel) at the frontal (F3, Fz, F4), the vertex (Cz) and
the parietal (P3 and P4) sites.
Despite the intersubject variability, grand average of EEG
activity and rectied theta and gamma components synchronized to ST1 and ST2 conrmed the signicance of the

Fig. 3. Top to bottom: EEG, rectied 37 and 3040 Hz, averaged


across subjects. Left column, F4; Right column, P4. The increase
of theta and gamma oscillations around ST1 and ST2, marked by
vertical lines at 0 s, is signicant (Wilcoxon rank test, P , 0:05).

above ndings (Wilcoxon rank test, P , 0:05). Fig. 3 illustrates these results for site F4 around ST1 and for site P4
around ST2. In accordance with [16], between ST1 and
movement end, the long lasting suppression of 810 Hz
and 1725 Hz activity at C3 0 , Cz and C4 0 , followed by
beta increase around the movement end, was observed
(not shown in Figures).
The PW and the preceding decrease of the variance did
not change when the negative trend was subtracted from the
data record (rst reconstructed component via singular
value decomposition [10]). Further, in averaged EEG to
an arbitrary chosen instance, no PW was observed, and no
signicant augmentation of theta and gamma oscillations
was detected (Wilcoxon rank test, P , 0:05). Hence, the
detected patterns are related to STs and are attributed to
higher frequency changes of the EEG signal.
To evaluate the variability of the instances of the ST
appearance, the time intervals between ST1, movement
onset, ST2 and the switch pressing were tested for randomness across the trials for each electrode and subject. The null
hypothesis for randomness was not rejected (CoxStuart test
for randomness, P 0:025). This indicates that no `idle'
EEG activity existed between the successive movements,
i.e. the subject is involved in the task during the whole
experimental session. One factor (electrode position)
ANOVA across the subjects showed no signicant difference of the mean intervals (P , 0:05).
Short-term transients indicate the exact time of changes
of the state of brain activity (in linear terms). As the experimental setup does not include any external cueing, STs
reect the aspects of pure voluntary movements. With
respect to the movement's time evolution and sites, the
PW shape and topography resemble P3 wave in the eventrelated potentials. Thus PW is like internally induced P3
that may be related to motor imagery and focusing of attention (a choice of the target and the way to reach it). The
association between PW and phase-locked theta oscillations
supports this notion, as both P3 and theta activity have been
related to episodic memory process responsible for orientation in space and time [7]. Similar proposition can be made
for parieto-central theta oscillations following ST2, when
the target is nearly reached and the switch is to be pressed.
Enhanced gamma oscillations appeared in parallel with
theta oscillations. The topography of the increased gamma
activity conrms similar results in [1,15]. Our results contribute to the role of the parietal cortex in a complex voluntary
movement organization [4,17]. It can be assumed that
gamma activity is probably related to increased information
transfer (high integration between the brain areas) in order
to nalize the task. Similar conclusions have been made for
synchronized gamma activity at distant cortical locations in
the processing of attended motor task [18]. Thus, the parallel between P3-like PW and gamma oscillations can be
interpreted as a key activity distinguishing intention to
start and to end the movement task and extends the ndings
when synchronization to movement onset is used [12,15].

D. Popivanov et al. / Neuroscience Letters 267 (1999) 58

High-resolution EEG techniques [3] could help for detailed


description of the brain localization of these activities.
This study was supported by the NFSR Grant No. 806/
1998-2001 (Bulgaria) and by the Project MAITE (1997-99)
between B.A.S. and A.S. of The Czech Republic. We thank
Drs. J. Yordanova, V. Kolev and R. Draganova for helpful
discussions and A. Tomova for technical assistance.
[1] Andrew, C. and Pfurtscheller, G., Event-related coherence
as a tool for studying dynamic interaction of brain regions.
Electroenceph. clin. Neurophysiol., 98 (1996) 144148.
[2] Deecke, L., Planning, preparation, execution and imagery of
voluntary action. Cogn. Brain Res., 3 (1996) 5964.
[3] Edlinger, G., Pfurtscheller, G., van Burik, M. and Wach, P.,
Post-movement EEG synchronization studied with different
high resolution methods. Brain Topogr., 10 (1997) 103113.
[4] Freund, H.J., Functional organization of human supplementary motor area and dorsolateral premotor cortex. In
H.O. Luders (Ed.), Advances in Neurology, Vol. 70, Supplementary Sensorimotor Area, Lippincott-Raven, New York,
1996, p. 263.
[5] Jansen, B.H., Time series analysis by means of linear
modelling. In J.P. Huston and R. Weikunat (Eds.), Techniques in the Behavioral and Neural Sciences, Vol. 5, Digital
Biosignal Processing, Elsevier, Amsterdam, 1991, pp. 157
180.
[6] Kaminsky, Y. and Krekule, I., Universal multifunctional IBM
PC I/O board for clinical examinations and experimental
research in neuroscience. Physiol. Res., 43 (1994) 193199.
[7] Klimesch, W., Schimke, H. and Schwaiger, J., Episodic and
semantic memory: an analysis in the EEG theta and alpha
band. Electroenceph. clin. Neurophysiol., 91 (1994) 428
441.
[8] Lange, D.H. and Inbar, G.F., A robust parametric estimator

[9]
[10]
[11]

[12]
[13]

[14]
[15]
[16]

[17]

[18]

for single-trial movement related brain potentials. IEEE


Trans. Bio-Med. Eng., 43 (1996) 341347.
Ljung, L., MATLAB System Identication Toolbox, The
MathWorks Inc, Natick, MA, 1993.
Mineva, A. and Popivanov, D., Method for single-trial readiness potential identication based on singular spectrum
analysis. J. Neurosci. Methods, 68 (1996) 9199.
Mineva, A., Krekule, I. and Popivanov, D., Short-term transients in the time-frequency dynamics of event-related EEG
detected by autoregressive models. In A. Prochazka, J.
Uhlir and P. Sovka (Eds.), Signal Analysis and Prediction
I, EURASIP, ICT Press, Prague, 1997, pp. 415417.
Pfurtscheller, G. and Neuper, C., Simultaneous EEG 10 Hz
desynchronization and 40 Hz synchronization during nger
movements. NeuroReport, 3 (1992) 10571060.
Pfurtscheller, G., Neuper, C., Schlogl, A. and Lugger, K.,
Separability of EEG signals recorded during right and left
motor imagery using adaptive autoregressive parameters.
IEEE Trans. Rehabil. Eng., 6 (1998) 316325.
Popivanov, D., Mineva, A. and Dushanova, J., Tracking EEG
signal dynamics during mental tasks. A combined linear/
nonlinear approach. IEEE Eng. Med. Biol., 17 (1998) 8995.
Salenius, S., Salmelin, R., Neuper, C., Pfurtscheller, G. and
Hari, R., Human cortical 40 Hz rhythm is closely related to
EMG rhythmicity. Neurosci. Lett., 213 (1996) 7578.
Stancak Jr, A. and Pfurtscheller, G., Desynchronization and
recovery of rhythms during brisk and slow self-paced nger
movements in man. Electroencheph. clin. Neurophysiol.,
196 (1995) 2124.
Urbano, A., Babiloni, C., Onorati, P. and Babiloni, F.,
Dynamic functional coupling of high resolution EEG potentials related to unilateral inernally triggered one-digit
movement. Electroenceph. clin. Neurophysiol., 106 (1998)
477487.
Yordanova, J., Kolev, V. and Demiralp, T., The phase-locking of auditory gamma band responses in humans is sensitive to task processing. NeuroReport, 8 (1997) 39994004.