Alpha oscillations as a correlate of trait anxiety

International Journal of Psychophysiology 53 (2004) 147 – 160 www.elsevier.com/locate/ijpsycho

Alpha oscillations as a correlate of trait anxiety Gennadij G. Knyazev *, Alexander N. Savostyanov, Evgenij A. Levin State Research Institute of Physiology, Siberian Branch of the Russian Academy of Medical Sciences, Timakova Street, 4, Novosibirsk 630117, Russia Received 29 September 2003; received in revised form 29 September 2003; accepted 24 March 2004 Available online 10 May 2004

Abstract The associations among psychometric measures of anxiety and depression and individually adjusted electroencephalogram (EEG) spectral power measures registered in resting condition and during experimental settings were investigated in 30 males aged 18 – 25 years. During all stages of registration, Taylor Manifest Anxiety and Spielberger state anxiety (SA) and trait anxiety (TA) scores were positively related to alpha and negatively to delta relative power with these relations being independent of cortical site. Within-subject estimate of the strength of reciprocal relationship between alpha and delta oscillations (alpha – delta anticorrelation, or ADA) was positively related to trait anxiety and depression. Three minutes after an alarming event (unexpected loud sound), a further increase of alpha power was observed. In low-anxiety subjects, this increase was mostly associated with fast alpha (alpha 3), whereas in high-anxiety ones, it was mainly linked to slow alpha (alpha 2). SA mediated relationship between TA and EEG power, while ADA and alpha band reactivity showed trait-like features being associated with TA even after accounting for SA. These findings are interpreted as an indication of higher vigilance and higher reactivity of alpha system in anxious individuals. D 2004 Elsevier B.V. All rights reserved. Keywords: EEG; Brain oscillations; Personality; Anxiety; Depression

1. Introduction Basar (1998), in his fundamental book on brain oscillations, points out that a great change is taking place in neuroscience due to the fact that brain scientists have recognized the importance of oscillatory phenomena and the functional electroencephalogram (EEG). A particularly important landmark in this book is the emphasis given to the alphas (i.e., distributed oscillatory processes in the 10-Hz frequency * Corresponding author. Tel.: +7-383-2-33-48-65; fax: +7-3832-32-42-54. E-mail address: [email protected] (G.G. Knyazev). 0167-8760/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.ijpsycho.2004.03.001

range). This great change has been fostered by development of new methodological approaches including ‘‘methods of thoughts’’ (Basar, 1998). Brain oscillations recorded in a form of spontaneous or evoked EEG could be considered as a kind of message carrying important information about intrinsic modes of brain activity. Given potential importance of this message, usage of EEG in psychology research seems inevitable. Knyazev and Slobodskaya (2003) proposed an evolutionary-based interpretation of brain oscillations relevant for research of EEG correlates of personality. Using as a starting point the concept of triune brain introduced by MacLean (1985), Knyazev and Slobod-

148

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

skaya (2003) suggested that delta, theta, and alpha oscillations reflect activities of three hierarchical philogenetic brain systems. Delta oscillations are linked with the most ancient system, which was dominant in reptilian brain. Theta oscillations dominate in lower mammals. Alpha oscillations are manifestations of activities of the newest system, which dominates in adult humans. The three hierarchical phylogenetic systems fulfill parallel processing, and their contribution to resulting behavior could differ in different individuals. The strength of reciprocal relationships between these systems and the relative prevalence of some oscillations over others relate to stable behavioral patterns relevant to personality and psychopathology. According to MacLean’s concept, the reptilian brain consists of the brainstem, cerebellum, and basal ganglia. Superimposed on this brain in the course of evolution is the brain of lower mammals, which added the limbic system. Finally, the third brain appeared in advanced mammals and primates in the form of added neocortex. It is important to emphasize that we do not equate exactly the three oscillatory systems with anatomical structures included in the three brains of the MacLean model; we just borrow his evolutionary idea. It should be kept in mind that the MacLean model is a theoretical construct, whereas the three oscillatory systems are empirical entities. So we could just look for existing data about evolution and distribution of these systems. Comparative studies by Bullock indicate that the most striking evolutionary puzzle is a general consistent difference in the power spectrum of EEG between all vertebrates and all invertebrates. This difference concerns the synchrony of slow waves ( < 50 Hz), which is dramatically higher in vertebrates. Invertebrates have much more obvious unit spiking than vertebrates, but much less relative amplitude of slow waves (Bullock, 1993). Among vertebrates, the degree of synchronization also increases during evolution. There is an evidence of less synchrony or more rapid coherence decline with distance in reptiles, amphibians, and fish than in mammals (Bullock, 1997). Beyond this, the power spectra look alike in all the vertebrates, falling quite steeply on each side of a maximum around 5– 15 Hz (Bullock, 1993). Oscillations of delta, theta, and alpha frequencies could be found in each vertebrate (Basar, 1998). But there is an important distinction between

reptiles, lower mammals, and humans in what frequency dominates in the scalp EEG. Alpha is the dominant frequency in adult humans, while theta dominates in the EEG of lower mammals (Klimesch, 1999) and delta in the reptilian EEG (Gaztelu et al., 1991; Gonzalez et al., 1999). That means that all three oscillatory systems were acquired early in the evolution of vertebrates, but they further developed with different rates. Development of delta system peaks in reptiles, theta in mammals, and alpha in primates. The three oscillatory systems do not have to exactly match anatomically the three brains of the MacLean model. They might be selectively distributed over the entire brain (Basar, 1998, 1999), although the main populations of neurons representing these systems must be located within respective brains. Following MacLean, one might speculate about physiological and behavioral correlates of these systems. The most ancient delta system deals with internally driven behavior oriented to acquisition of biologically important goals, such as survival, physical maintenance, dominance, and mating. Among environmental signals, this system mostly recognizes those relevant to biologically important motives or current goals. Interestingly, randomly distributed splashes of delta activity have been repeatedly registered during presentation of target stimuli in P300 experimental paradigm (Basar, 1998). Increase in delta power might be expected in states of increased biological motivation (e.g., sexual). Indeed, Schutter and van Honk (in press) have recently shown that in healthy male volunteers, administration of testosterone, which presumably enhances sexual motivation, significantly increases the delta power. In children, prevalence of delta oscillations has been shown to relate to parent, teacher, and self-ratings of conduct disorder (Knyazev et al., 2002b, 2003), which is also in line with the above reasoning. According to MacLean, the reptilian brain is active, even in deep sleep, since it controls autonomic functions, such as breathing and heartbeat. That explains why delta is the most salient rhythm during the slow-wave sleep. Theta system operates in close conjunction with the delta system, but it is linked with more flexible behavior regulation, which implies the matching of internal drives with acquired during lifetime experience. Alpha system is engaged in perception and recognition of environmental patterns. Increase of

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

alpha activity during sensory stimulation is linked to specific sensory cortices (Basar, 1998). Both theta and alpha oscillations have been associated with memory processes, but Klimesch (1999) emphasizes crucial distinction between the two in what kind of memory they are linked with. While theta is associated with contextual memory, alpha is engaged in semantic memory processes. These two kinds of memory differ not only in temporal characteristics but also mostly in their content. Contextual memory stores sensual images. All mammals including humans share this kind of memory. Semantic memory is a store of knowledge, which is enormously developed in human beings. Prevalence of alpha oscillations and enhanced anticorrelation between alpha and delta bands has been shown to relate to behavioral inhibition (BI) and trait anxiety (Knyazev et al., 2002a; Knyazev and Slobodskaya, 2003). Why is alpha activity enhanced in anxious individuals? Gray and other researchers of behavioral inhibition have recognized long ago that activity of the behavioral inhibition system (BIS) is linked with permanent scanning of environment in search of potentially threatening factors. Particularly, novelty is also a BIS-relevant stimulus as it signifies a potential threat (Gray, 1987). It is clear then that in anxious individuals, a system dealing with perception and recognition of environmental patterns should be more prepared for information processing especially in a new or ‘‘strange’’ environment, such as psychophysiological laboratory. One may argue that alpha power is in opposite relation to the activity level; therefore, if the alpha system is expected to be more active in anxious individuals, that would show itself in less alpha power. Such train of thought made Eysenck and Eysenck (1985) predict that extraverts should have more alpha power, since their theory posits that extraverts are less aroused than introverts. But most meta-analysts and reviewers of the relevant literature note equivocal relations between extraversion and EEG measures (Gale, 1983; O’Gorman, 1984; Bartussek, 1984). It is scarcely surprising because intraindividual regularities may not be automatically extended to interindividual differences. When different states of an individual are considered, alpha oscillations are indeed more synchronous during relatively inactive states, but that does not necessarily mean that an individual with more alpha power is

149

more relaxed than an individual with less alpha power when both of them are measured in similar conditions. For example, it is well known that heart rate tends to increase with increasing muscle activity. Therefore, it could be taken as a measure of muscle activity. Now if we compare two Ss and find that one of them has slower heart rate than another one, we might conclude that the current level of muscle activity is higher in the latter subject. But actually, if we measure heart rate in a stayer even during moderate muscle activity (e.g., mounting a stairway), it might be found to be slower than the heart rate of an individual with low fitness measured in a resting state. Besides, it should be kept in mind that EEG power reflects the number of neurons that discharge synchronously (Klimesch, 1999). Therefore, different and sometimes conflicting reasons may lead to the same result (i.e., increased alpha power). As has been noted elsewhere (Knyazev et al., in press a,b), the total power of a given rhythm in the site of registration depends on the mean power of oscillation and the number of currently active individual oscillators belonging to this oscillatory system, on a degree of their synchrony and on reciprocal relationships between oscillatory systems. Short-lasting changes of alpha power may mostly depend on a degree of alpha synchrony, whereas long-lasting changes may reflect a changing number of active alpha oscillators. In our opinion, interindividual differences of alpha power should be considered as differences in alpha system preparedness for information processing, with higher power being an indicator of higher vigilance. There are several recent reports of alpha synchronization in tasks associated with the maintenance of attention during anticipation of visual events (e.g., Fernandez et al., 1998; Orekhova et al., 2001). There is another important question. Personality characteristics such as trait anxiety are stable over time and should have some structural basis. Brain oscillatory activity reflected in EEG is very labile and strongly depends on an individual’s state. How can these two be compared? Yet, test – retest reliability of resting EEG has been shown to be comparable with that of personality questionnaires (Kondacs and Szabo, 1999), thus demonstrating trait-like properties of the labile EEG signal. We consider that possible correlations between the two domains should reflect correlation between the individuals’ traits and states.

150

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

For example, in the same environment, an anxious individual should probably show more state anxiety. Therefore, we suggest that the relationship between EEG power and trait anxiety is mediated by state anxiety. The present study aims to replicate previous findings and to overcome some their limitations. The adult sample of studies by Knyazev et al. (2002a, 2003) and Knyazev and Slobodskaya (2003) was mainly females. EEG recordings were made only in resting conditions and from a limited number (six) of head sites. In the present study, we aimed to test whether association between the personality trait of behavioral inhibition and EEG found previously in resting females could be reproduced in males in different stages of a psychophysiological experiment, particularly under condition of unexpected arousing stimulus (alarm). We also sought to investigate the cortical specificity of EEG – personality relationship and to test whether the relationship between trait anxiety and EEG is mediated by state anxiety.

2. Materials and methods 2.1. Subjects For the purpose of the present study, data from a previously published study (Savostyanov et al., 2002) were used. That study aimed to investigate the effect of trait and state anxiety on parameters of averaged evoked potentials (AEPs; namely, amplitude and latency of P300 component) during manipulation on subjects’ attention to external stimuli. Subjects were 31 right-handed nonpsychology students, males, Caucasians, aged 18– 25 years (mean = 21.2; S.D. = 3.5). One subject was discarded due to inconsistency in psychometric data (MA = 1.1 S.D.; SA= + 1.8 S.D.). All participants gave consent to completing the selfreport questionnaires and the psychophysiological protocols. 2.2. Instruments and procedures Russian versions of the Spielberger State Trait Anxiety Inventory (STAI; Spielberger et al., 1970; Hanin, 1989), Taylor Manifest Anxiety Scale (MAS; Taylor, 1953), and Beck Depression Inventory (BDI;

Beck et al., 1974) were used as psychometric measures of anxiety and depression. Annet (1970) test was used to evaluate handedness. These were filled out just before the experimental procedure. Physiological measures were obtained in the afternoon. Each participant was seated comfortably in a reclined armchair with eyes closed within a soundinsulated dimly lit chamber. The participants were asked to minimize their movements during the recording. Prior to recordings, the individual thresholds of acoustical sensitivity were measured for each subject. Experimental procedure consisted of several sessions, during four of which EEG were recorded. After 1 min of baseline recording, (baseline) subjects were presented with 30 acoustic stimuli (tone 1000 Hz, 500 ms, 50 dB above individual threshold, with interstimulus interval of 1.28 s) for AEP registration. Then three previously tape-recorded words (conclusion, finger, and porch) were presented 20 times in triplets with the interstimulus interval of 10 s. This procedure aimed to extinguish attention to external stimuli. Part of the Ss (eight) was asked to count the words and mark every fourth of them by pressing a button of a device, thus preventing decline of attention. Subsequent repeated-measures ANOVA showed that these Ss did not differ from the rest of the group on averaged EEG measures during the experiment with all main effects of the grouping variable and its interaction with experimental procedure and trait anxiety being nonsignificant. Therefore, for the purposes of the present study, both groups were pooled together. EEG was recorded throughout the presentation of words (word presentation). Then Ss were informed that the word presentation is over. Next, the abovedescribed procedure of AEP registration was repeated. Background EEG was recorded during 1 min (background before alarm) and, unexpectedly, a loud sound (tone 1000 Hz, 500 ms, 100 dB) was given. After a delay of about 1.5 min, registration of AEP was repeated and then EEG was recorded for the last time (alarm session). For more details of the experimental procedure, see Savostyanov et al. (2002). 2.3. Psychophysiological recording EEGs were recorded using a 32-channel PC-based system via silver –silver chloride electrodes. A mid-

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

forehead electrode was the ground. The electrode resistance was maintained below 5 kV. The signals were amplified with a multichannel biosignal amplifier with bandpass 0.05 –70 Hz, 6 dB per octave, and continuously digitised at 300 Hz. The electrodes were placed into 28 head sites according to the International 10 – 20 system and referred to linkedear electrode. The horizontal and vertical electrooculographies (EOGs) were registered simultaneously. Data contaminated with muscle or blink artifact was discarded offline. 2.4. Psychophysiological data reduction Fast Fourier transforms (FFTs) of artifact-free EEG chunks 6.8 s in duration were then performed. For each subject, 4 chunks of baseline record, 18 chunks during word presentation, 4 chunks of relaxation, and 4 chunks of activation periods were analyzed. There are different suggestions about EEG rhythm boundaries. Traditionally, regular activity below 4 Hz is considered as delta rhythm, activity between about 4 and 7.5 Hz as theta, and 7.5 – 12.5 Hz as alpha (Klimesch, 1999). Klimesch suggests that broad-band analyses should be interpreted with caution because alpha and theta frequencies vary considerably as a function of age, neurological disease, brain volume, etc. He recommends the use of individual frequency bands distinguishing three alpha bands. We have shown previously that the use of individually adjusted bands substantially increases the size and significance of correlations between EEG and personality measures (Knyazev et al., 2002a, 2003). Therefore, in this study, we also used alpha frequency, averaged over all leads and all baseline epochs, as the anchor to adjust frequency bands individually for each subject (Klimesch, 1999). There are, however, some questionable points linked with this method. First, since the alpha 1 subband in Klimesch’s subdivision actually falls into what is traditionally considered as theta diapason, it should be treated as alpha subband with caution (Pfurtscheller and Lopes da Silva, 1999). Next, the alpha frequency could be defined in terms of peak or gravity frequency within the traditional range of 7.5– 12.5 Hz. According to Klimesch, if there are multiple peaks in the alpha range, gravity frequency is the better estimate of alpha. Basing on this, we previously used the gravity frequency as the estimate of alpha. It

151

is clear that the two estimates will coincide only if there is a sharp and ideally symmetrical peak in the alpha range, which is rarely the case. For example, if the peak is skewed, the gravity frequency will shift toward a more gentle slope. In this study, we used the gravity frequency [ f (i)], which was calculated as the weighted sum of spectral estimates, divided by alpha power: f (i) = S[a( f )xf ]/S[a( f )]. Power spectral estimates at frequency f are denoted as a( f ). The index of summation is in the range 7.5– 12.5 Hz. To compare the two methods of alpha frequency determination, we also used peak alpha frequency. In 5 of 30 Ss, there was no alpha peak. These Ss were lower on all anxiety measures, although only difference on SA reached significance (T = 2.2, p = 0.036). Then, four frequency bands with a width of 2 Hz were defined in relation to f (i) determined either as gravity or peak frequency. The frequency bands obtained were termed: theta [ f (i) 6 to f (i) 4]; alpha 1 [ f (i) 4 to f (i) 2]; alpha 2 [ f (i) 2 to f (i)], and alpha 3 [ f (i) to f (i) + 2]. The delta band was defined as approximately 0.2 Hz to f (i) 6. The beta band was defined as f (i) + 2 to 30 Hz. The absolute power measures usually used are influenced by factors such as electrical resistance at the site of electrode placement and skull thickness. Use of relative power measures should diminish data variability and increase the probability of EEG –personality relationships emerging. On the other hand, calculating relative spectral powers adds a part of the variability of the most powerful bands to other spectral bands, resulting in the danger of artificial covariation. Therefore, in the present study, natural logarithms of absolute spectral powers were calculated in addition to relative spectral power measures. Logarithms are conventionally used to normalize the distribution of absolute power measures. Relative spectral powers were calculated for every electrode location and each EEG band as a band spectral power divided by total spectral power. In parts of analyses, the power measures for each frequency band were averaged across all electrode placements. Within-subject estimates of oscillatory systems relationships (OSRs) were evaluated as described previously (Knyazev and Slobodskaya, 2003). Specifically, natural logarithms of mean delta, theta, and alpha (i.e., alpha 1 + alpha 2 + alpha 3) absolute spectral powers across all electrode locations were calcu-

152

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

lated for every 30 chunks for each S. Then a principal components analysis was performed for each S on the correlations matrix of these three variables with three components extracted. As has been shown previously (Knyazev and Slobodskaya, 2003), as a rule, all three natural logarithms of mean delta, theta, and alpha absolute spectral powers had high positive loadings on the first component with negative relationships emerging in the second and third components. In some Ss, however, negative loadings might appear in the first component. In our previous study (Knyazev and Slobodskaya, 2003), there were just six such Ss (out of 146) and we excluded these Ss from the consequent analyses. In this study, however, probably because of a smaller number of available for each S EEG chunks, half of the sample showed negative loadings on the first component. These Ss did not differ from the rest of the sample, neither on psychometric nor on mean EEG power variables. To rearrange factor loadings in these Ss, we used orthogonal target rotation of previously extracted unrotated components by means of the statistical software program CEFA 1.10, developed by Browne et al. (2002). This was done following the recommended procedures by establishing a binary target of 1s and 0s to specify a hypothesized factor loading. Specifically, all three variables (i.e., delta, theta, and alpha absolute spectral powers) were specified as having high positive loadings (1s) on the first component and low loadings (0s) on the remaining two components. This approach was effective in all Ss with deviant factor structure. That is, after the target rotation in all these Ss, the three variables showed high positive loadings on the first component, whereas negative relationships showed up in the remaining two factors. The first factor loadings were used to calculate the factor scores. That was done by summing up the z-transformed values of the three variables multiplied by respective factor loadings. In Ss with normal factor structure, target rotation was not performed and the scores of the first unrotated factor were used. Partial correlations between the three variables were then calculated by controlling for the first factor scores. All correlations must be negative, which only might be achieved with a reasonably large sample of EEG chunks. One S showed a positive correlation between delta and theta power, and this S was excluded from consequent analyses. The inverted correlation coefficients were used as

individual measures of OSR, namely, alpha – delta anticorrelation (ADA), alpha – theta anticorrelation (ATA), and theta –delta anticorrelation (TDA).

3. Results There are no normative data on trait anxiety (TA) measures in the Russian population. In this sample, mean (S.D.) TA (STAI) was 41.3 (9.6). In a reasonably large sample (n = 307) of students and their relatives, we Knyazev et al. (in press b) have recently collected data on a number of psychometric measures including TA. In that sample, mean (S.D.) TA was 41.8 (9.9). Therefore, the mean level of TA in the present study sample could be considered as average. Table 1 shows cross-correlations of psychometric measures. All these correlations are positive and highly significant. The two measures of trait anxiety (TA and MA) correlate at 0.92, signifying that they actually capture the same construct. They are also strongly correlated with depression scores. State anxiety, on the other hand, shows only moderate correlations with trait anxiety and depression. Fig. 1 shows grand average of EEG spectrum for Ss with high (0.5 S.D.) and low ( 0.5 S.D.) MA averaged over all leads and all chunks within baseline session. It is evident that anxious Ss have more alpha power. Interestingly, there is a separate peak in slow alpha range in anxious, but not in nonanxious Ss. The means (S.D.) of alpha frequency were 10.43 (0.78) for peak frequency and 10.13 (0.24) for gravity frequency. These two measures were significantly correlated (r = 0.87, p < 0.001). Both estimates negatively correlated with anxiety measures, but only correlation of peak frequency with SA reached significance (r = 0.43, p = 0.037). Repeated-measures ANOVA was used to test for trait anxiety  cortical zone interaction separately for Table 1 Cross-correlations of psychometric measures (n = 30)

(1) (2) (3) (4)

Trait anxiety (STAI) State anxiety (STAI) Manifest anxiety (Taylor) Depression (Beck)

1

2

3

– 0.60 0.92 0.83

– 0.62 0.62

– 0.79

All correlations are significant at p < 0.001.

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

153

Fig. 1. EEG spectrum averaged over all leads and all chunks within baseline session. Grand averages for Ss with high (0.5 S.D.) and low ( S.D.) MA.

0.5

each frequency band and every state. Relative power measures in different cortical areas were treated as repeated measures. To preserve statistical power, Taylor Manifest Anxiety Scale scores were entered as a covariate. Table 2 shows that for all states and frequency bands, MA  cortical zone interaction was not significant (although for alpha 2, they approached significance in the baseline and word presentation sessions). The main effect of trait anxiety was signif-

icant for relative delta, theta, and alpha 2 power. Given nonsignificant effects of MA  cortical zone interaction, all consequent analyses were performed using EEG measures averaged across all head sites. Correlations of psychometric measures with relative power measures, averaged across all head sites and all states, are presented at Table 3. Mean relative delta power negatively correlated with MA and SA. Correlation with TA was marginal ( p = 0.059). Mean

Table 2 Eta squared for main effects of MA on EEG bands’ relative power (before oblique stroke) and its interactions with the head site (after oblique stroke)

Table 3 Correlations of psychometric measures with averages across all head sites and all states relative power measures (n = 30)

Delta Theta Alpha 1 Alpha 2 Alpha 3 Beta – gamma

Baseline

Words

Background

Alarm

0.23( )/ 0.06 0.17( )/ 0.02 0.00/0.02 0.32(+)/ 0.11 0.06/0.01 0.04/0.08

0.18( )/ 0.10 0.11/0.02

0.06/0.07 0.02/0.03

0.01/0.01 0.24(+)/ 0.11 0.02/0.03 0.06/0.05

0.02/0.01 0.14(+)/ 0.08 0.00/0.02 0.01/0.07

0.19( )/ 0.06 0.20( )/ 0.02 0.01/0.01 0.29(+)/ 0.10 0.03/0.02 0.07/0.05

If significant at p < 0.05, the direction of association is shown in parentheses.

Trait anxiety Delta Theta Alpha 1 Alpha 2 Alpha 3 Beta TDA ADA ATA

0.35 0.18 0.08 0.39* 0.21 0.19 0.22 0.43* 0.10

State anxiety 0.45* 0.29 0.01 0.53** 0.36 0.03 0.11 0.27 0.29

Manifest anxiety 0.44* 0.16 0.19 0.48** 0.24 0.25 0.24 0.42* 0.06

Depression 0.27 0.24 0.06 0.22 0.28 0.12 0.13 0.55** 0.29

* Correlation is significant at the 0.05 level (two-tailed). ** Correlation is significant at the 0.01 level (two-tailed).

154

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

relative alpha 2 power was positively related to all anxiety measures and alpha 3 mean relative power was marginally related to SA ( p = 0.053). Absolute alpha power also positively correlated with anxiety measures and the size of correlations was similar to the one presented in the Table 3. Among OSR measures, only ADA showed significant correlations, being positively related to anxiety and depression. Interestingly, contrary to power measures, ADA correlated most strongly with depression and trait anxiety, while its correlation with state anxiety was not significant. Next, we examined the role of state anxiety as a mediator of relationships between trait anxiety and EEG power measures. Following the general guidelines of Baron and Kenny (1986), to establish evidence for the proposed mediation model, it would be necessary to meet three conditions. First, trait anxiety must be significantly associated with state anxiety. Second, trait anxiety must be associated with the outcome, in this case EEG power. Third, the association between trait anxiety and the outcome should be reduced when controlling for the influence of state anxiety. Complete mediation would be indicated if the association between trait and state anxiety, and between state anxiety and EEG power, were significant, but the previously significant association between trait anxiety and EEG power became nonsignificant after controlling for the influence of state anxiety. Mediation was tested using hierarchical regression. Only mediation effects of state anxiety on relationships between Taylor’s MA scale and delta and alpha 2 relative power were tested because all other associations did not fulfill the above criteria. After controlling for state anxiety, neither delta (b = 0.23, T = 1.2, p = 0.24) nor alpha 2 (b = 0.21, T = 1.22, p = 0.233) relative power was not anymore a significant predictor of Taylor Manifest Anxiety. It might be concluded, therefore, that the relationship between trait anxiety and EEG relative power is mediated by state anxiety. Repeated-measures ANOVA was used to test the effects of TA and experimental conditions (ECs) separately for each frequency band. Absolute power averaged across all cortical areas was used as a repeated measure. EC factor included five levels, viz. baseline, beginning of word presentation (first 10 words), ending of word presentation (last 10

words), background before alarm, and alarm. TA grouping variable with two levels (below and above median) was entered as a between-subject factor. For delta, theta, and alpha 1 absolute power, neither main effect nor interaction was significant. For alpha 2, the effect of experimental condition ( F = 5.196, df = 3, p = 0.006) and TA  EC interaction ( F = 4.593, df = 3, p = 0.010) emerged as significant. The same was true for alpha 3 band ( F = 7.035, df = 3, p = 0.001 and F = 3.35, df = 3, p = 0.034 for EC and TA  EC, respectively). For high-frequency (beta) bands, only the main effect of TA was significant ( F = 5.069, df = 1, p = 0.032), indicating that absolute power of highfrequency oscillations was uniformly higher in anxious subjects across all experimental conditions. According to Klimesch (1999), the alpha 2 and alpha 3 subbands are functionally different. The former is associated with nonspecific attention, whereas the latter is linked with semantic memory. One may expect, therefore, that the two subbands would react differently to different experimental conditions. To test that, three repeated-measures ANOVAs were performed with two bands (alpha 2 vs. alpha 3) and five ECs as repeated measures; TA, or MA, or SA grouping variable was used as a between-subject factor. The main effects of band and EC and TA  EC EC interaction were significant, but all other effects were not. We repeated the same analyses with alpha power estimates determined in terms of alpha peak frequency. In this case, the interaction band  EC was significant ( F = 4.67, df = 4, p =0.009) as well as the interaction band  EC  SA ( F = 4.76, df = 4, p = 0.008). As Fig. 2 shows, both in low- and highanxiety Ss, the beginning of word presentation produces an increase of alpha 3 power, whereas alpha 2 power decreases in low-anxiety Ss and does not change in high-anxiety Ss. Towards the end of word presentation, both alpha 2 and alpha 3 powers decrease in low-anxiety Ss. In high-anxiety Ss, both of them do not change substantially and even tend to increase. When Ss were informed that the word presentation is over, in low-anxiety Ss, alpha 2 power begins to increase, whereas alpha 3 continues to decrease. Again there are no substantial changes in high-anxiety Ss. Finally, alarm produces increase of both alpha 2 and alpha 3 power in both low- and highanxiety Ss, but in low-anxiety Ss, alpha 3 increases more sharply compared to alpha 2, whereas in high-

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

Fig. 2. Effect of experimental condition on mean alpha 2 and alpha 3 power in subjects with low (top) and high (bottom) state anxiety (SA). Bsl—baseline; Words1—beginning of word presentation; Words2—ending of word presentation; Bkgr—background before alarm.

anxiety Ss, the pattern is opposite—here the increase of alpha 2 clearly prevails. To obtain the ratio of specific vs. unspecific alpha activation in each experimental condition, alpha 3 absolute power was divided by respective alpha 2 power and these ratios for the five experimental conditions were used as repeated measures with SA grouping variable entered as a between-subject factor. Both EC ( F = 4.49, df = 4, p = 0.010) and EC  SA interaction ( F = 4.34, df = 4, p = 0.012) were significant. Fig. 3 shows that in high-anxiety Ss, in the beginning of word presentation, the alpha 3/alpha 2 ratio sharply increases. It stays almost constant during the next two recordings and decreases after alarming

155

sound. In low-anxiety Ss, initial increase of this ratio gives place to its sharp decrease, but alarm again produces its increase. Tests of within-subject repeated contrasts showed that effect of SA  EC interaction was significant for the ending of word presentation vs. background before alarm ( F = 5.17, df = 1, p = 0.033) and background before alarm vs. alarm ( F = 5.12, df = 1, p = 0.034). To obtain measures of reactivity of different spectral bands following alarming stimulation, differences of absolute power measures between the two states (alarm minus background before alarm) were calculated. Reactivity of alpha 2 band was positively related to the initial alpha 2 level during background before alarm session (r = 0.58, p = 0.001). It was also associated with TA (r = 0.48, p = 0.006), SA (r = 0.43, p = 0.016), MA (r = 0.58, p = 0.001), and depression (r = 0.38, p = 0.040). Association with MA remained significant even after controlling for the initial alpha 2 power (r = 0.43, p = 0.022). Reactivity of alpha 3 band was associated with MA (r = 0.39, p = 0.028) and marginally with TA (r = 0.34, p = 0.062) and depression (r = 0.33, p = 0.083). After controlling for SA, correlation of MA with alpha 2 reactivity remained significant (r = 0.46, p = 0.011), while correlation between SA and alpha 2 reactivity lost significance after controlling for MA (r = 0.20, p = 0.294). To reveal associations between anxiety, alpha 2 reactivity, and initial power of alpha 2 oscillations, a

Fig. 3. Effect of experimental condition on alpha 3/alpha 2 ratio in subjects with low and high state anxiety (SA). Bsl—baseline; Words1—beginning of word presentation; Words2—ending of word presentation; Bkgr—background before alarm.

156

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

grouping variable with four levels was constructed. The first group included 18 Ss with both initial level of alpha 2 power during background before alarm session and its reactivity after alarm being below median. The second group consisted of four Ss with low initial alpha 2 power but high reactivity. Four Ss of the third group had high initial power but low reactivity. At last, the fourth group included five Ss with high initial power and high reactivity. One-way ANOVA showed that this grouping variable had significant effect on MA [ F(3,30) = 3.76, p = 0.022] and SA [ F(3,30) = 5.12, p = 0.006]. Post-hoc LSD tests showed that the fourth group with both initial power and reactivity above median scored significantly higher than all other groups on state anxiety. This group also scored higher on MA than both groups with low reactivity (first and third groups), but it did not differ significantly from the second group with low initial level but high reactivity (Fig. 4). Finally, we checked for cortical specificity in the effect of alarming stimulation on alpha 2 power in individuals with different levels of TA. Repeatedmeasures ANOVA with EC (background before alarm vs. alarm) and cortical derivate (28 levels) as withinsubject factors and TA as a covariate showed that only main effects of EC [ F(1,30) = 10.00, p = 0.004] and TA [ F(1,30) = 4.59, p = 0.041] and their interaction

Fig. 4. The z-scores of state anxiety (SA) and manifest anxiety (MA) in Ss with different levels of initial alpha 2 power (before alarm) and alpha 2 reactivity (difference of alpha 2 power after and before alarm). lil-lr—Low initial level and low reactivity; lil-hr—low initial level and high reactivity; hil-lr—high initial level and low reactivity; hil-hr—high initial level and high reactivity.

[ F(1,30) = 11.75, p = 0.002] were significant, whereas effects of derivates and all its interactions were not.

4. Discussion The present study confirmed and extended our previous findings (Knyazev et al., 2002a, 2003; Knyazev and Slobodskaya, 2003) showing that positive association of behavioral inhibition (i.e., trait anxiety) with relative alpha power and its negative association with relative delta power are evident not only in females but also in males. These associations were observed not only in resting condition but also in all other experimental sessions. They lost significance only in the session preceding alarming stimulation due to the fact that alpha power increased in low-anxiety Ss but not in high-anxiety Ss, thus smoothing differences between the two groups. Gale (1983) once noted that the level of arousal during experimental settings is crucial for revealing EEG – personality relationships. Knyazev et al. (2003), following Gale, also suggested that not only a level of arousal but also a level of state anxiety is crucial, implying that increase in state anxiety should deteriorate the relationship observed in resting conditions. The present study has shown, however, that it is not so. Actually, transition from the state with presumably lower arousal and state anxiety to the state with heightened arousal (alarm) made the EEG– TA relationship even more salient. Knyazev et al. (in press a) have shown that at least for six cortical sites, the EEG – personality relationship does not depend on the cortical locus. This study confirmed it for a larger set of cortical points. In anxious Ss, prevalence of alpha and decrease of relative delta power, as well as alpha power increase during transition to a more activated state, were not restricted to specific cortical areas implying that global properties of oscillatory systems underlie individual differences in behavioral inhibition. For the OSR estimates, the previously shown relations (Knyazev and Slobodskaya, 2003) were confirmed. The strength of reciprocal relationship between alpha and delta oscillations was positively related to trait anxiety. Robinson (1999, 2000, 2001) was the first one who decomposed the betweensubject covariance among AEP components filtered

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

in delta, theta, and alpha ranges into positive and negative parts. He speculated that the negative relationship between alpha and delta oscillations is a reflection of negative relationship between thalamus and reticular formation. Knyazev and Slobodskaya (2003) applied this approach to spontaneous EEG and to within-subject domain and putatively identified the alpha – delta anticorrelation with the strength of descending inhibition from the more evolutionary advanced alpha to the more ‘‘ancient’’ delta system. It should be emphasized, however, that the meaning of positive and negative covariance between delta, theta, and alpha waves and underlying mechanisms might not be identical for evoked interindividual and spontaneous intraindividual variability. The intraindividual OSR measures repeatedly demonstrated their validity and relevance to some physiological mechanisms underlying personality differences. The use of targeted rotation effectively eliminates the problem with deviant factor structure in some subjects. Two points are important when these measures are used. First, slow waves (delta) must be evaluated reliably. That implies that the EEG epochs must be of appropriate length (at least 4 s). The shorter the epoch, the less is the reliability of slow-wave estimates. Second, the number of EEG chunks available for factor analysis and calculation of anticorrelations must be sufficient. The size of correlation between ADA and trait anxiety observed in the present study should be considered as very substantial given inevitable errors linked with measuring of both estimates. For example, self-reports of anxiety might be biased when they are obtained in young males, particularly in a more traditional culture, such as the Russian culture (Triandis, 1995), because anxiety is not ‘‘appropriate’’ gender role behaviour for males. Indeed, we found out that exclusion of just one subject dramatically increased the size and significance of correlations between ADA and psychometric variables (r = 0.66, p < 0.001, r = 0.51, p = 0.005, r = 0.50, p = 0.007, for correlation of ADA with depression, TA, and MA, respectively). This subject scored 1.4 S.D. below the mean for this sample on both depression and TA, probably underestimating or concealing his anxiety symptoms. All psychometric measures yielded similar results in terms of their associations with EEG measures,

157

although the magnitude of these associations was different. It is noteworthy that whereas EEG power measures were strongly associated with SA, which mediated their relationship with TA, measures of reactivity and OSR showed trait-like features. Thus, correlation of alpha 2 reactivity with SA lost significance after controlling for TA. But it should be noted that due to small sample size, it is dangerous to draw definite conclusions from the absence of a significant correlation when the statistical power of tests is low. As for the alpha – delta anticorrelation, this estimate did not show significant association with SA, whereas its associations with TA, MA ,and depression were significant. Particularly interesting is a moderate association of ADA with depression. Given our hypothesis about motivational aspects of delta system function, this finding hints at possible physiological mechanisms underlying depressive disorders. Two methods of determining alpha frequency (peak and gravity) yielded similar results, but it seems that peak frequency divides alpha 2 and alpha 3 subbands more accurately. Indeed, subband-specific changes during experimental settings were significant for peak-derived—but not for gravity-derived—estimates. Since gravity frequency shifts toward a more salient slope, it tends to smooth differences between alpha 2 and alpha 3. On the weak side of the peak method, it is not applicable in Ss with no peak. The different dynamics of alpha 2 and alpha 3 changes during transition through experimental sessions is in keeping with Klimesch’s (1999) notion about these subbands’ functions. At the onset of word presentation, all Ss show increase of alpha 3 power. This signifies increased activity of specific system dealing with semantic memory and word recognition. Further presentation of the same words results in loss of interest and sharp decline of alpha 3 power in lowanxiety Ss. Those high on state anxiety, however, do not cease to attend. The most salient difference between the two groups reveals itself during alarm. Whereas low-anxiety Ss tend to react to unexpected events by increase of specific attention (increase of alpha 3/alpha 2 ratio), those high on state anxiety react by increase of unspecific attention (decrease of alpha 3/alpha 2 ratio). It might be speculated that increase of alpha 3/alpha 2 ratio reflects an attempt to understand the meaning of a happening. That would imply use of knowledge and semantic memory. Contrary to that,

158

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

decrease of this ratio could be interpreted as a sign of alertness and preparedness to fight or flight. In all these conditions, increase of alpha activity is associated with increase of alpha power. Is this not in contradiction with traditional ideas of alpha desynchronization as a typical reaction to activation? First, we must remember that all observed changes took place during minutes after a provoking event. For example, EEG recording in alarm session has been done approximately 3 min after the alarming event when actual desynchronization was over and postreaction adjustment occurred. In experiments with event-related desynchronization, time intervals are usually much shorter. Moreover, whereas alpha desynchronization is unequivocally considered as a sign of activation, contradictory interpretations of alpha synchronization could be met in literature. For example, Pfurtscheller and Lopes da Silva (1999) consider alpha synchronization as a means of deactivation of correspondent networks. But Basar (1998, 1999) presents ample and convincing data indicating that in cross-modality experiments both in cats and humans, the differences between responses to adequate/inadequate stimulation were marked in the alpha frequency range, hinting at a possible special role of the alpha response in primary sensory processing. The evoked alpha oscillations usually have a damped oscillatory character lasting for about 300 ms and are restricted to primary sensory cortices. It is important to emphasize that evoked alpha enhancement is actually reorganization and phase locking of ongoing alpha activity (Basar, 1998, 1999). That means that a sensory event does not engage in activity additional alpha oscillators, which were inactive before, but instead it phase-locks and synchronizes ongoing activity. Thus, short-lasting time-locked alpha synchronization in specific cortices during sensory stimulation could be considered a specific phenomenon linked with sensory processing. An anonymous referee of this paper suggested that increased alpha power in anxious Ss is associated with increased activity in thalamocortical circuits. We also acknowledge the importance of distinguishing between the activity of alpha oscillators and their synchrony (Knyazev et al., in press a). But the impact of these two processes is difficult to separate. For example, transition from inactive to an active state may increase the number of currently active alpha oscillators, but simultaneously, it may decrease their syn-

chrony. As a result, the registered power of alpha oscillations may increase, decrease, or remain seemingly unchanged, depending on relative impact of these processes. Further research is needed for the evaluation of the relative contribution of these two processes. Owing to extensive studies by Basar as well as other authors, a considerable body of knowledge has been accumulated, indicating that depending on background activity, different reactions of EEG bands could be observed. Specifically, high-amplitude spontaneous alpha activity coincided with alpha blocking, while low-amplitude alpha preceded evoked potentials (EPs) of high amplitude (Basar, 1998). According to the concept proposed by Basar (1998), the ongoing EEG determines (controls) evoked activity. This signifies that through the maintenance of higher or lower power of specific oscillations, the brain could be prepared or predisposed to specific patterns of responses. Klimesch (1999) states that the reactivity in band power can be predicted from the amount of absolute power as measured during a resting state. Considering the reactivity of alpha and theta bands, he notes that large alpha power in the reference interval is associated with a large amount of desynchronization during task performance. The opposite holds true for the theta band. Here, small reference power is related to a large amount of synchronization. He concludes that the most reactive individuals would show in resting condition significantly more power in the alpha band but less power in the theta band. This is exactly the case when we compare anxious individuals with less anxious ones. Higher reactivity of anxious individuals is well established using heart rate and skin conductance as indicators of arousal (e.g., Garrada et al., 1991; Mezzacappa et al., 1997). We may conclude that enhanced alpha and relatively decreased slow-wave power in anxious Ss indicate predisposition to high reactivity of brain oscillatory systems. It might be hypothesized, therefore, that within the domain of individual differences, higher alpha power reflects readiness to mobilize resources in anticipation of some difficult task or threat. Thus, according to Klimesch (1999), more alpha power in a resting state should predict better performance in some cognitive tasks. Specifically, high background alpha power (especially with prevalence of slow alpha) may be

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

an evidence of predisposition to react to external stimuli by general unspecific activation, which is characteristic for anxious endophenotype. As we can see from the present study, this ‘‘tuning’’ of alpha system is state-dependent. In anxious Ss, it increases with appearance of unexpected or threatening stimuli. Note that the mode of reaction to unexpected stimuli by increase in alpha power seems more peculiar to anxious individuals than background alpha power. The former depends more on trait and the latter on state. Particularly the group with low initial level of alpha but high reactivity showed a low level of SA but high level of TA. It might be deduced that these Ss, in spite of having anxious endophenotypes, did not feel anxiety at the onset of experiment but loud sound was sufficient to evoke in them anxious reaction. Gray (1987) linked anxiety with activity of the limbic system, emphasizing the role of hippocampal theta. The hypothesized link between behavioral inhibition and the limbic system is based on a large amount of animal research, which shows that the limbic system, particularly the amygdala, is responsive to unfamiliar events and seems to be necessary for the acquisition of conditioned freezing reactions to the unconditioned stimulus of shock (i.e., reactions associated with behavioural inhibition; e.g., Davis, 2000). But one recent study has demonstrated that, in primates, fear-related or anxiety responses characteristic of temperament, stable and present from early in life, are not mediated by the amygdala but are mainly linked with cortical regions, while the amygdala has an important role in mediating initial responses to fearful stimuli (Kalin et al., 2001). Knyazev et al. (2003) supposed that the hypotheses by Gray linking behavioural inhibition and theta oscillations, which does not show itself in resting conditions, might emerge under conditions of heightened emotion. This study showed that it is not the case. It is noteworthy that reactivity in alpha band was positively associated with initial alpha power and that there were more anxious Ss who simultaneously demonstrated both the highest initial alpha power and the highest increase in alpha power. Interestingly, it has been shown that in children with high behavioral inhibition, the law of initial values does not hold for heart rate reactivity. That is, these children show both higher resting heart rate and higher heart rate

159

increase during stimulation (Snidman, 1989; Slobodskaya, 1989). Concordant with this evidence, the present study findings confirm that even in resting condition, anxious individuals tend to tune their systems to a state of higher readiness, and this tuning dramatically increases following a warning signal from the environment.

Acknowledgements This study was supported by a grant from the Russian Foundation for Basic Research (RFBR; no. 03-06-80058-a). We are grateful to D.A. Savostyanova and N.V. Dmitrienko for assistance with data collection and to E.R. Slobodskaya for helpful comments.

References A classification of hand preference by association analysis. Br. J. Psychol. 61, 303 – 321. Baron, R.M., Kenny, D.A., 1986. The moderator – mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J. Pers. Soc. Psychol. 51, 1173 – 1182. Bartussek, D., 1984. Extraversion und EEG: Ein Forschungsparadigma in der Sackgasse? In: Amelang, M., Ahrens, H.J. (Eds.), Brennpunkte der Personlichkeitsforschung. Hogrefe, Gottingen, pp. 157 – 189. Basar, E., 1998. Brain Function and Oscillations: I. Brain Oscillations, Principles and Approaches. Springer, Berlin. Basar, E., 1999. Brain Function and Oscillations: II. Integrative Brain Function, Neurophysiology and Cognitive Processes. Springer, Berlin. Beck, A.T., Rial, W.Y., Rickets, K., 1974. Short form of Depression Inventory: cross-validation. Psychol. Rep. 34, 1184 – 1186. Browne, M.W., Cudeck, R., Tateneni, K., Mels, G., 2002. CEFA: Comprehensive Exploratory Factor Analysis, Version 1.10, computer software and manual (http://quantrm2.psy.ohio-state. edu/browne/). Bullock, T.H., 1993. How Do Brains Work? Birkhouser, Boston. Bullock, T.H., 1997. Signals and signs in the nervous system: the dynamic anatomy of electrical activity is probably informationrich. Proc. Natl. Acad. Sci. U. S. A. 94, 1 – 6. Davis, M., 2000. The role of the amygdala in conditioned and unconditioned fear and anxiety. The Amygdala: A Functional Analysis. Oxford Univ. Press, Oxford, pp. 213 – 287. Eysenck, H.J., Eysenck, M.W., 1985. Personality and Individual Differences: A Natural Science Approach. Plenum, New York. Fernandez, T., Harmony, Th., Silva, J., Galan, L., Diaz-Comas, L., Bosch, J., Rodriguez, M., Fernandez-Bouzas, A., Yanez, G.,

160

G.G. Knyazev et al. / International Journal of Psychophysiology 53 (2004) 147–160

Taylor, J.A., Marosi, E., 1998. Relationship of specific EEG frequencies at specific brain areas with performance. NeuroReport 9, 3681 – 3687. Gale, A., 1983. Electroencephalographic studies of extraversion – introversion: a case study in the psychophysiology of individual differences. Pers. Individ. Differ. 4, 371 – 380. Garrada, M.E., Connel, J., Taylor, D.C., 1991. Psychophysiologic anomalies in children with emotional and conduct disorders. Psychol. Med. 21, 947 – 957. Gaztelu, J.M., Garcia-Austt, E., Bullock, T.H., 1991. Electrocorticograms of hippocampal and dorsal cortex of two reptiles: comparison with possible mammalian homologs. Brain Behav. Evol. 37 (3), 144 – 160. Gonzalez, J., Gamundi, A., Rial, R., Nicolau, M.C., De Vera, L., Pereda, E., 1999. Nonlinear, fractal, and spectral analysis of the EEG of lizard, Gallotia galloti. Am. J. Physiol. 277, R86 – R93 (Regul. Integr. Comp. Physiol., 46). Gray, J.A., 1987. Perspectives on anxiety and impulsivity: a commentary. J. Res. Pers. 21, 493 – 509. Hanin, Y.L., 1989. Cross-cultural perspectives of the individual differences diagnostic. Vopr. Psihol. 4, 118 – 125. Kalin, N.H., Shelton, S.E., Davidson, R.J., Kelley, A.E., 2001. The primate amygdala mediates acute fear but not the behavioral and physiological components of anxious temperament. J. Neurosci. 21 (6), 2067 – 2074. Klimesch, W., 1999. EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. Brain Res. Rev. 29, 169 – 195. Knyazev, G.G., Slobodskaya, H.R., 2003. Personality trait of behavioral inhibition is associated with oscillatory systems reciprocal relationships. Int. J. Psychophysiol. 48, 247 – 261. Knyazev, G.G., Slobodskaya, H.R., Wilson, G.D., 2002a. Psychophysiological correlates of behavioural inhibition and activation. Pers. Individ. Differ. 33, 647 – 660. Knyazev, G.G., Slobodskaya, H.R., Aftanas, L.I., Savina, N.N., 2002b. EEG correlates of the emotional and behavioural problems in schoolchildren. Physiol. Cheloveka (Hum. Physiol.) 28, 16 – 22. Knyazev, G.G., Slobodskaya, H.R., Safronova, M.V., Sorokin, O.V., Goodman, R., Wilson, G.D., 2003. Personality, psychopathology and brain oscillations. Pers. Individ. Differ. 35 (6), 1331 – 1349. Knyazev, G.G., Slobodskaya, H.R., Wilson, G.D., in press a. Personality and brain oscillations: developmental aspects. In: Advances in Psychology Research. Nova Science Publishers, New York. Knyazev, G.G., Slobodskaya, H.R., Wilson, G.D., in press b. Comparison of construct validity of the Gray – Wilson personality questionnaire and the BIS/BAS scales. Pers. Individ. Differ.

Kondacs, A., Szabo, M., 1999. Long-term intra-individual variability of the background EEG in normals. Clin. Neurophysiol. 110, 1708 – 1716. MacLean, P.D., 1985. Brain evolution relating to family, play, and the separation call. Arch. Gen. Psychiatry 42, 405 – 417. Mezzacappa, E., Tremblay, R.E., Kindlon, D., Saul, J.P., Arseneault, L., Seguin, J., Pihl, R.O., Earls, F., 1997. Anxiety, antisocial behaviour, and heart rate regulation in adolescent males. J. Child Psychol. Psychiatry 38, 457 – 469. O’Gorman, J.G., 1984. Extraversion and the EEG: I. An evaluation of Gale’s hypothesis. Biol. Psychol. 19, 95 – 112. Orekhova, E.V., Stroganova, T.A., Posikera, I.N., 2001. Alpha activity as an index of cortical inhibition during sustained internally controlled attention in infants. Clin. Neurophysiol. 112, 740 – 749. Pfurtscheller, G., Lopes da Silva, F.H., 1999. Event-related EEG/ MEG synchronization and desynchronization: basic principles. Clin. Neurophysiol. 110, 1842 – 1857. Robinson, D.L., 1999. The technical, neurological and psychological significance of ‘alpha,’ ‘delta’ and ‘theta’ waves confounded in EEG evoked potentials: a study of peak latencies. Clin. Neurophysiol. 110, 1427 – 1434. Robinson, D.L., 2000. The technical, neurological and psychological significance of ‘alpha,’ ‘delta’ and ‘theta’ waves confounded in EEG evoked potentials: a study of peak amplitudes. Pers. Individ. Differ. 28, 673 – 693. Robinson, D.L., 2001. How brain arousal systems determine different temperament types and the major dimensions of personality. Pers. Individ. Differ. 31, 1233 – 1259. Savostyanov, A.N., Illjutchenok, R.Ju., Savostyanova, D.A., 2002. Dynamics of acoustical evoked potentials on a background of habituation to verbal stimulus with the people, who have the different individual anxiety level. Acta Biol. Hung. 53, 575 – 587. Schutter, D.J.L.G., van Honk, J., in press. Decoupling of midfrontal delta – beta oscillations after testosterone administration. Int. J. Psychophysiol. Slobodskaya, H.R., 1989. Behaviour and autonomic nervous system activity of toddlers in the nursery. PhD Thesis, Novosibirsk. Snidman, N., 1989. Behavioral inhibition and sympathetic influence on the cardiovascular system. Perspectives on Behavioral Inhibition. Chicago Univ. Press, pp. 51 – 69. Spielberger, C.D., Gorsuch, R.L., Lushene, R.E., 1970. Manual for the State – Trait Anxiety Inventory. Consulting Psychologists Press, Palo Alto, CA. Taylor, J.A., 1953. A personality scale of manifest anxiety. J. Abnorm. Soc. Psychol. 48, 285 – 290. Triandis, H., 1995. Individualism and Collectivism. Westview, Boulder, CO.