Influence of lateralized sensorimotor and neuropsychological activities on electroencephalographic spectral power

200

Electroencephalography and clinical Neurophysiology , 1990, 75: 200-206 Elsevier Scientific Publishers Ireland, Ltd.

EEG 02338

Influence of lateralized sensorimotor and neuropsychological activities on electroencephalographic spectral power B. De Toffol, A. Autret, S. Markabi and S. Roux Clinique Neurologique, C H U Bretonneau, 37000 Tours (France) (Accepted for publication: 20 June 1989)

Summary In order to test the effect of 'lateralized' sensorimotor and neuropsychological activities on EEG spectral power (P), we recorded EEGs over the right and left central regions with 3 different derivations C5P3/C6P4, C5Cz/C6Cz, and C5 ears-linked (EL)/C6EL in 14 young, right-handed men who underwent ten 2 min sequences including 4 during rest (3 with eyes closed (EC), 1 with eyes open (EO)), and 6 during tasks reputed to involve preferentially the left (Le) or the right (Ri) hemisphere, i.e., pure left and pure right motor activity of the hand (EC), pure neuropsychological tasks consisting in lexical followed by spatial form analysis and finally mixed (neuropsychological and motor) EC tasks consisting in writing followed by left hand object recognition. Three spectral parameters P, log P and asymmetry index AI = (P Ri - P Le)/(P Ri + P Le) were calculated in 5 frequency bands 0, al, a2, a3 and

/~1. We observed a relationship between task complexity and P reduction on both hemispheres which was greater during rest, less during motor activity and least during mixed motor and neuropsychological activities; during EC activities more specifically involving the left hemisphere, only the comparison between rest and mixed sequences was significant. During EC activities involving the right hemisphere, both motor and mixed sequences were significantly different from rest sequences. In addition, during the EO sequences, P was usually greater than during rest. As compared to rest conditions, the AI increased during sequences preferentially involving the left hemisphere and decreased during preferentially right hemisphere activities. The number of rhythms involved varied according to the lead type and the parameter chosen. Therefore, this report supports the hypothesis that in the frequency bands studied, P decreases according to the complexity of the cerebral activity, being even lower on the hemisphere directly implicated in the task. Key words: EEG spectral analysis; Sensorimotor activity; Neuropsychological activity

The localizationist theory of cerebral function can be studied by EEG spectral analysis if we accept the idea that electrical activity varies according to the different tasks accomplished by the cerebral hemispheres. In order to test this hypothesis, sensitive electrophysiological parameters and discriminatory tasks must be found. Derivation choice, electrophysiological parameters and statistical analysis have been reviewed by Gevins and Schaffer (1980) and by Beaumont

Correspondence to: Dr. B. De Toffol, Clinique Neurologique, CHU Bretonneau, 37000 Tours (France).

(1983), and a number of studies have shown the variations in interhemispheric spectral power during sensorimotor activities in normal subjects (Storm van Leeuwen et al. 1978; Van Huffelen et al. 1984; Pfurtscheller 1988). Usually, sensorimotor activity reduces contralateral spectral power (Autret et al. 1985) and considerably hampers the evaluation of the effects of neuropsychological activities (Gevins et al. 1980). The importance of environmental factors and attention has been studied by Ray and Cole (1985). In this study, we compared the effects of pure sensorimotor, pure neuropsychological, and mixed (neuropsychological associated with sensorimotor)

0013-4649/90/$03.50 © 1990 Elsevier Scientific Publishers Ireland, Ltd.

EFFECT OF LATERALIZEDSENSORIMOTOR ACTIVITY ON EEG tasks which supposedly specifically implicate one of the cerebral hemispheres. We used parameters taken from spectral analysis of the EEG recorded symmetrically from the right and left central region (C5 or C6), using 3 different derivations. We also tested the hypothesis whereby: (1) spectral power decreases in relation to the complexity of cerebral activity, and (2) the decrease is greater over the hemisphere directly implicated in the task studied.

Material and methods

Fourteen young (21-24 years old), right-handed (Oldfield's test 1971), males placed in a comfortable armchair in a partially darkened and soundproofed room underwent 10 sequences each lasting 2 min. Commands were communicated verbally between each session and the absence of eye movement was verified by an electro-oculogram. Rest periods, 3 with the eyes closed (REC) and one with the eyes open (REO), alternated with periods of activity. Eyes closed (EC) activity sessions consisted in a sequence of hand movements (manipulation of clay placed in the subject's hands) using the right hand (RiMEC), and another using the left hand (LeMEC); writing a memorized text with the right hand (WEC), left-hand recognition and classification of a series of cardboard objects having different sizes and shapes and placed in a box situated on the subject's left (OEC). Eyes open (EO) activity sessions consisted in a test requiring choosing a synonym (Binois and Pichot 1974) (SYN EO) and a test involving rotating flat geometric figures (Thurstone and Thurstone 1951) (ROT EO), to correspond with a test picture placed 50 cm in front of the subject. The progression and performance of each test were controlled. All 10 sessions were arranged according to a latin square in order to eliminate any effect related to the passing order. A bipolar derivation (C6P4/C5P3), a vertex reference derivation (C6Cz/C5Cz) and a reference derivation connecting both ears (C6-ears linked ( E L ) / C 5 E L ) were used simultaneously. The EEG was amplified using an Alvar polygraph (3 dB; 0.5-2.5 kHz). Sampling was accomplished in real

201

time at a frequency of 100 Hz using an analog-digital converter coupled to a Minc 11/03 and stored on a disk. Before sampling, the signal was filtered by anti-aliasing filters having a cut-off frequency of 30 Hz and an attenuation of 24 d B / o c t a v e . Processing was deferred. Successive elementary 2.56 sec spectra were calculated and each analysis period was filtered by temporal multiplication, using a Hanning window. Forty-eight elementary spectra were used to establish a definitive average spectrum. We considered spectra with a power greater than 200 /~V2/Hz in the frequency band 0-1 Hz to be artefacts, and consequently rejected them. Accordingly, 3 parameters were defined: P (average power in # V 2 / H z during a session, log P, and the asymmetry index AI

PRi- PLe PRi+PLe

The frequency bandwidths studied were: 0, 3.91-7.03 Hz; a 1, 7.03-8.98 Hz; or2, 8.98-10.94 Hz; a 3, 10.94-12.89 Hz; fla, 12.89-19.92 Hz.

Statistical analysis We compared the different parameter values between sessions by a 1-way ANOVA of log P and AI (repeated measures of unidimensional variance analysis). We made 3 different comparisons: (1) rest EC sessions amongst one another in order to test parameter stability; (2) between EC sessions i n order to look for the existence of significant differences between active and rest sessions and between sensorimotor and mixed sessions; (3) between EO sessions so that the role of neuropsychological tasks could be determined. When a significant intersequence difference was discovered ( P < 0.05), we compared each active sequence and rest sequence on the one hand, and the motor sequences to the neuropsychological sequences supposedly involving the same hemisphere on the other hand, i.e., R i M E C / W E C and L e M E C / O E C , by a Student's t test (df= 13). The log P and AI results are presented separately using untransformed power values (P), but all significances are based on log P values.

202

B. D E T O F F O L E T A L .

Results

(1) Rest EC sequences No significant difference was found between the rest sequences irrespective of the rhythm, the derivations or the parameters studied (ANOVA between the 3 rest EC sequences for log P and AI). (2) Tasks EC Results of the comparison of the EC sequences (REC, RiMEC, LeMEC, OEC, and WEC) for log P are reported in the left part of Table I, according to derivations and rhythm. (a) The average spectral power values for each hemisphere, each rhythm, and each derivation in which a significant difference ( P < 0.05) appeared during log P ANOVA are reported in Table II. The results obtained in C6Cz/C5Cz are shown in Fig. 1. (aa) With C6Cz/C5Cz, a significant intersequence difference exists for 4 of the 5 rhythms (0, Otl, a3, /~1)" The P values decrease in a constant order: P is stronger during rest over the 2 hemispheres, 'intermediate' during sensorimotor sequences (RiMEC, LeMEC) and weaker during mixed tasks (OEC and WEC). OEC and WEC tasks produce significantly lower power than during rest 14 out of 16 times while RiMEC and LeMEC only produce significantly lower power 3 times: two during LeMEC and one during RiMEC. The comparison between RiMEC and WEC on the one hand and LeMEC and OEC on the other shows that the power is significantly stronger during LeMEC than during OEC for a 1 ( P < 0.05) and t31 ( P < 0.01) rhythms in both hemispheres.

There is no significant difference between RiMEC and WEC. (ab) With C 6 E L / C 5 E L , the same intersequence differences are found. The P value is stronger during rest over the 2 hemispheres and is significantly lower during mixed tasks 7 out of 16 times. P comparisons during sensorimotor tasks and during mixed tasks only show a stronger power for fll during the LeMEC test on the right hemisphere. With C6P4/C5P3, a significant intersequence difference only appears during a 3 and fll rhythms. The power is stronger on the two hemispheres during rest while mixed neuropsychological tasks induce a significant decrease 5 times out of 8. When comparing LeMEC and OEC, the power is significantly stronger for a 3 and fll rhythms on the two hemispheres ( P < 0.05) during LeMEC. There is no significant difference between RiMEC and WEC. (b) Comparison of the AI only shows a significant intersequence difference for al, a 2 and fll rhythms in C6P4/C5P3 and a 2 and a 3 in C6Cz/C5Cz: these results are illustrated in Table III. For the 75 total measurements (3 derivations × 5 sequences × 5 rhythms), a significant difference was found 25 times. With C6C4/C5P3, AI is always positive (power is stronger on the right hemisphere). Tasks involving the left hemisphere significantly increase the index as compared to rest ( P < 0.01, rhythms a~ and a 2 during WEC and rhythm a2 during RiMEC). There is a significant difference between RiMEC and WEC (the index is higher on the right for a 1 during WEC, P < 0.05) but none between LeMEC and OEC.

TABLE I Intersequencedifferencesin EEG power(P),l-way Rhythm

ANOVAIog P valuesfor EC(leftpart)

EC

a n d E O ( f i g h t p ~ t ) sequences.

EO

C6P4/C5P3

C6Cz/C5Cz

C6EL/C5EL

C6P4/C5P3

C6Cz/C5Cz

C6EL/C5EL

0

NS

***

*

***

NS

NS

a1 a2

NS NS

*** NS

* NS

** ***

** *

* *

a3 fll

** *

*** ***

* **

*** NS

** NS

NS NS

* P < 0.05; * * P < 0.01; * * P < 0.001; NS, non-significant.

E F F E C T OF L A T E R A L I Z E D S E N S O R I M O T O R A C T I V I T Y ON EEG

203

TABLE II Spectral power values for derivations and rhythms having a significant A N O V A difference for the log P parameter (sequences EC) and comparison of the spectral power for each active sequence with respect to rest (Student's t test). REC, rest EC; Le, left; Ri, right; M, motor; O, object recognition; W, writing. Rhythm C 6 C z / C 5 C z REC 0

LeMEC

C6EL/C5EL RiMEC

OEC

WEC

REC

7.01

6.92

6.46

6.42

10.12

RiMEC

OEC

WEC

7.78 NS 8.26 NS

8.03 NS 8.59 NS

7.55

*

NS

9.34 NS 9.84 NS

9.12 NS 10 NS

9.03 NS 9.86 NS

9 NS 8.08 NS

6.45 * 5.43 **

8.42 6,53 NS 8.66 7,14 NS

6.77 NS 7.34 NS

6.42 ** 6.92 ***

2.57 NS 1.97

3.31

2.69 NS 2.56 NS

2.4 ** 2.4

Ri

8.34

Le

8.82

7.22 NS

7.16 NS

6.73 NS

6.46 *

Ri

13.01

Le

12.2

10.22 NS 9.42 NS

9.54 NS 8.61 NS

7.6 ** 6.21 *

7.49 * 6.61 *

a 3 Ri

8.8

6.89 NS 7.72 6.16 NS

7.02 NS 5.84 NS

5.52 *** 5.55 **

3.17

2.43 NS 2.07

2.21 *** 1.92

a~

Le

fll

Ri Le

2.37 NS 2.77 2.14 *

*

*

*

* *

*

*

LeMEC

C6P4/C5P3

7.98 NS 9.36 8.48 NS

14.11 14.1

2,62 NS 3.22 2.51 NS

*

*

REC

LeMEC

R i MEC

OEC

WEC

7.14 * 6.69 *

10.51

9.47 NS 10.26 NS

9.69 NS 10.04 NS

7.26 *** 8.2 **

8.42 NS 8.68 *

2.85 NS 2.56 NS

2.35

1.98 NS 1.97 1.75 NS

2.05 NS 1.7 NS

1.81 *** 1.67

2.24 NS 1.87 NS

*

7.95 NS

11.21

*

*

* P < 0.05; * * P < 0.01; * * * P < 0.001; NS, non-significant.

With C6Cz/C5Cz, the index becomes negative during tasks implicating the right hemisphere (a 2 for LeMEC and a 2 and a 3 for OEC). OEC decreases the index as compared to rest ( P < 0.01). There is a significant difference ( P < 0.01) in a 3 between RiMEC and WEC and between LeMEC and OEC (power is stronger on the left during OEC, P < 0.01). Finally, whatever the rhythm, a significant difference exists ( P < 0 . 0 1 ) in C6P4/C5P3 and C 6 C z / C 5 C z between OEC and WEC: power is stronger on the right 5 times out of 5 during writing (or weaker on the left in the case of object recognition).

(3) Effects of EO tasks The results of the comparisons of the EO sequences (REO, SYNEO, ROTEO) for log P are reported in the right part of Table I. (a) The P values for each hemisphere, each rhythm, and each derivation which showed a significant difference ( P < 0.05) during the ANOVA of log P are reported in Table IV.

With C6Cz/C5Cz, there is a significant intersequence difference for the a t, ot2 and a 3 rhythms. The power is always stronger on the two hemispheres during rest and. reaches the significance

TA B LE III Asymmetry index (AI) values in derivations and rhythms in which EC sequences are different for AI A N O V A ( P < 0.05) and comparisons of the AI of each active sequence with respect to rest. Abbreviations: see Table II. REC

LeMeC

R i MEC

OEC

WEC

0.071 NS 0.028 NS 0.052 NS

0.07 NS 0.113 ** 0.061 NS

0.027 NS 0.077 NS 0.029 NS

0.129 ** 0.135 ** 0.11 NS

0.018

-- 0 . 0 6

0.005

-- 0 . 0 7

0.082

0.024

NS 0.012 NS

NS 0.051 NS

NS - 0.061

NS 0.037 NS

C6P4/C5P3 a1

0.0284

a2

0.061

fll

0.054

C6Cz/C5Cz Ot2

a3

* *

204

B. D E T O F F O L ET AL.

TABLE IV Spectral power values for derivations and rhythms having a significant A N O V A difference for the log P parameter (sequences EO) and comparison of the spectral power for each active sequence with respect to rest (Student's t test). REO, rest EO; ROT, rotation; SYN, synonym. Rhythm

0

a1

a2

C6P4/C5P3

C6Cz/C5Cz

EO

ROT

SYN

R

2.77

L

2.59

3.18 NS 3.02

3.58 ** 3.45

R

5.14

L

4.10

2.61 NS 2.51

2.84 NS 2.65

•

*

R

9.06

4.75 •

L

7.32

5.33

R

3.49

L

3.41

REO

ROT

6.59

4.82 * 5.12

7.01

SYN

5.01 * 5.27

NS

11.65

***

10.10 *

ROT

SYN

5.26 NS 5.94 NS

5.84 * 6.14 NS

11.46

8.19

6.23

11.91

9.69

6.66

7.46 7.85

9.22 **

12.76

10.97 NS

10.6 **

2.50 • 2.39

2.52 * 2.25

5.13

4.36 * 4.18

4.53 NS 3.89

•

*

**

*

4.92

R EO

*

4.66 **

•

aa

4.67

C6EL/C5EL

* P < 0.05; ** P < 0.01; * * * P < 0.001; NS, non-significant.

threshold 9 times out of 12. With C6EL/C5EL, there is a significant intersequence difference only for the a I and a 2 rhythms. The power is always higher during rest on the two hemispheres and the differences reach a significant level 5 times out of 8. With C6P4/C5P3, there is a significant intersequence difference for the 0, a 1, a 2 and a 3 rhythms. For the a rhythms, the power is stronger on both hemispheres during rest; in contrast, the 0 power is weaker during rest than during the neuropsychological tasks. In 3 of 4 cases, the difference is significant (P < 0.05) during ROT EO in the left hemisphere and ( P < 0.01) during SYN EO. (b) The AI is significantly greater (P < 0.05) during SYN EO tasks than during rest or during ROT EO tasks for the a 3 rhythm with C6Cz/C5Cz.

Discussion

The precautions in methodology suggested by Gevins and Schaffer (1980) were followed: only completely right-handed subjects were chosen,

symmetrical tasks were compared, performance and accomplishment were controlled, ~ and f12 rhythms were not considered since they are susceptible to artefacts from ocular and muscular movements and, finally, the spontaneous EEG activity recorded on paper was controlled during the entire study. During task accomplishment, we avoided allowing lateral gaze movements on the part of our subjects, since their EEG effects have been previously demonstrated (Autret et al. 1985). We chose tasks which supposedly specifically implicate one of the two hemispheres, in accordance with a lesional model: RiMEC, WEC and SYN EO sequences respectively correspond to fighthand paralysis, agraphia and alexia secondary to left hemisphere lesions, while LeMEC, OEC and ROT EO respectively correspond to left-hand paralysis and visuo-spatial defects secondary to right hemisphere lesions (Gardner 1975). The tests were chosen with the aim of establishing an 'activity level' hierarchy in order to study the effect of simple motor activity, simple neuropsychological activity and the effect of these two simultaneous activities during mixed sequences.

E F F E C T OF L A T E R A L I Z E D S E N S O R I M O T O R ACTIVITY O N E E G

P

p.v2/Hz

8.t . 4.

(zl

o~3

131

2.o

:

": *

0,

REC

LeMEC

RiMEC

OEC

WEC

Hemisphere

Fig. 1. Spectral power values in each hemisphere for rhythms 8, Otl, a 3 and fll in C 6 C z / C 5 C z during REC, LeMEC, RiMEC, O E C and W E C sequences. The significant comparisons between active and rest sequences in each hemisphere and during a given rhythm are represented by an asterisk. * P < 0.05, • * P < 0.01, * * * P < 0.001.

The neuropsychological tasks included a measurable operative activity whereas, in the rest sequences, no particular instructions were given. Moreover, we only used activity sequences which required environmental observation (intake tasks) (Ray and Cole 1985). We employed the complementary parameters P and log P and an asymmetry index, which gave information on 'activity level' and 'lateralization' respectively, since Beaumont (1983) did not recommend using an

205

asymmetry index alone. Using a 1-way ANOVA allowed us to assume that normal distribution existed: the statistical distribution of the spectral values approaches gaussian distribution after logarithmic transformation (Gasser et al. 1982). AI is a standard parameter used in spectral analysis; we verified that it varied precisely with log P Ri/log P Le. Using ANOVA avoids relying on a multiple univariate test because we only used Student's t test when there were significant differences in ANOVA. Our study is not comparable to that of Amochaev and Salamy (1979) who only used a single asymmetry parameter in comparing 7 neuropsychological tasks in 6 subjects in order to test the results of stability between subjects according to different derivations while repeating the sequences, or to that of Davidson et al. (1980) who used a single asymmetry parameter in testing the frequency bands 9-11 Hz in 27 right-handed subjects during 4 neuropsychological tasks. Our report clearly shows, as Pfurtscheller demonstrated in 1988, that derivation choice influences results in relation to the parameter selected and the experimental conditions during EO and EC. Thus, the reference derivation has a greater number of significant intersequence differences for EC 'activity levels' measured by log P than for the bipolar derivation. C6EL/C5EL is less sensitive than the two others for detecting intersequence differences with respect to log P and AI in the EO sequences. In 15 comparisons, C6EL/C5EL detected 4 significant differences versus 9 for C6Cz/C5Cz and 11 for C6P4/C5P3. Finally, a 3 rhythm intersequence differences were always present, whatever the derivation or the sequence type. In the vertex and bipolar derivations the variations in spectral power confirm our initial hypothesis: for the rhythms studied the spectral power on the two hemispheres decreases in relation to cerebral activity in the following order: rest, sensorimotor activity and mixed activity (sensorimotor and neuropsychological). This is in agreement with Adrian and Matthews' classic model (1934) which states that 'ct' is inversely proportional to the level of mental activity. Activity sequences constantly have a lower power than

206

rest sequences and mixed sequences have a lower power than corresponding sensorimotor ones. This last finding supports the hypothesis that neuropsychological activity has a specific influence which must be differentiated from the sensorimotor component of the task. When comparing motor activity to rest, spectral power differences were statistically significant only 3 of 40 times. This finding may be due to electrode placement since motor activity is best studied by using the C3-C4 derivation (Storm van Leeuwen et al. 1978). These results also confirm our second hypothesis whereby there is a greater fall in spectral activity on the hemisphere which is supposedly involved in the task. The asymmetry index during rest always shows greater spectral power on the right hemisphere and the significant variations in the index are always toward an accentuation of this asymmetry on the right in tasks involving the right side of the body (greater involvement of the left hemisphere) and a lessening of this asymmetry and occasionally a negative index (in C6Cz/C5Cz) during tasks involving the left side of the body (the power is stronger on the left hemisphere than on the right). In C6P4/C5P3 for a a rhythm, the power is significantly greater on the right during WEC with respect to RiMEC and in C 6 C z / C 5 C z and is significantly stronger on the left for the ct3 rhythm during OEC with respect to LeMEC. Therefore, there is a hierarchy in the asymmetry index according to the degree of hemispheric task involvement. Finally, whatever the derivation or the rhythm, comparisons of the AI during the right and left neuropsychological sequences always exhibit a decrease in power in the hemisphere which is supposed to be the more active during the task, as was suggested previously by Davidson and Ehrlichman (1980).

References Adrian, E.D. and Matthews, B.H.C. The Berger rhythm: potential changes from the occipital lobe in man. Brain, 1934, 57: 355-399.

B. DE TOFFOL ET AL. Amochaev, A. and Salamy, A. Stability of EEG laterality effects. Psychophysiology, 1979, 16: 242-246. Autret, A., Auvert, L., Laffont, F. and Larmande, P. Electroencephalographic spectral power and lateralized motor activities. Electroenceph. clin. Neurophysiol., 1985, 60: 228-236. Beaumont, J.G. The EEG and task performance: a tutorial review. In: A.W.K. Gaillard and W. Ritter (Eds.), Tutorials in ERP Research: Endogenous Components. North-Holland Publ., Amsterdam, 1983: 385-406. Binois, R. et Pichot, P. Test de Vocabulaire. Centre de Psychologie Appliqure, Paris, 1974. Davidson, R.J. and Ehrlichman, H. Lateralized cognitive processes and the electroencephalogram. Science, 1980, 207: 1005-1006. Davidson, R.J., Taylor, N., Saron, C. and Stenger, M. Individual differences and task effects in EEG measures of hemispheric activation. I. Effects of reference electrode. Psychophysiology, 1980, 17: 9a, 311. Gardner, E. Fundamentals of Neurology. The Forebrain Cerebral Specialization, 6th Edition. Saunders, Philadelphia, PA, 1975: 420-430. Gasser, T., B~icher, P. and M~cks, J. Transformations towards the normal distribution of broad-band spectral parameters of the EEG. Electroenceph. clin. Neurophysiol., 1982, 53: 119-124. Gevins, A.S. and Schaffer, R.E. A critical review of electroencephalographic (EEG) correlates of higher cortical functions. CRC Crit. Rev. Bioeng., 1980, 4: 113-164. Gevins, A.S., Doyle, J.C., Schaffer, R.E., Callaway, E. and Yeager, C. Lateralized cognitive processes and the electroencephalogram. Science, 1980, 207: 1006-1007. Oldfield, R.C. Edinburgh handedness inventory. Neuropsychologia, 1971, 1: 97-113. Pfurtscheller, G. Mapping of event-related desyncbronization and type of derivation. Electroenceph. clin. Neurophysiol., 1988, 70: 190-193. Ray, W.J. and Cole, H.W. EEG alpha activity reflects attentional demands, and beta activity reflects emotional and cognitive processes. Science, 1985, 228: 750-752. Storm van Leeuwen, W., Wieneke, G., Spoelstra, P. and Versteeg, H. Lack of bilateral coherence of mu rhythm. Electroenceph. clin. Neurophysiol., 1978, 44: 140-146. Thurstone, L.L. and Thurstone, T.G. Chicago Tests of Primary Mental Abilities. Edition Franqaise. Centre de Psychologie Appliqu~e, Paris, 1951. Van Huffelen, A.C., Poortvliet, D.C.J. and Van der Wulp, C.J.M. Quantitative electroencephatography in cerebral ischemia. Detection of abnormalities in 'normal' EEGs. In: G. Pfurtscheller, E.J. Jonkman and F.H. Lopes da Silva (Eds.), Brain Ischemia: Quantitative EEG and Imaging Techniques. Progress in Brain Research, Vol. 62. Elsevier, Amsterdam, 1984: 3-28.