Antero-Posterior EEG Spectral Power Gradient as a Correlate of Extraversion and Behavioral Inhibition

INTRODUCTION

The general tendency in the research of brain oscillations is to study effects averaged both in the within- (i.e., across consecutive trials) and the between- (i.e., across individuals) subject domains. Such approach helps to reveal major effects that are stable across time and subjects. However, this approach largely precludes systematic investigation of time-varying patterns and individual differences in observed effects. Analysis of variability of short-term spectral patterns (SPs) shows that a limited number of SP types may describe the human electroencephalogram (EEG) accurately [1, 2]. However, only about half of all SP types are functionally active (i.e., their appearance depends on the functional state of the brain) [1]. Individual-specific SP characteristics show trait-like properties being relatively stable over long periods of time [3]. This concurs with rapidly accumulating evidence showing heritability of some important characteristics of brain oscillations. Twin studies indicate that a considerable part of inter-individual variation in EEG variables could be attributed to genes [4]. The heritability estimates are about 79% for EEG alpha power (somewhat lower for other frequency bands), 81% for alpha peak frequency, 60% for P300 amplitude and EEG coherence measures [4-6]. It has been also shown that about 60% of the variance in the temporal structure of amplitude fluctuations in spontaneous EEG oscillations could be attributed to genetic factors [7].

One characteristic of the short-term SPs that is relatively less studied is the temporal variability and individual differences in the cortical distribution of spectral power in different frequency bands. It is well established that different cortical areas participate in different functional processes. On the other hand, it is increasingly becoming clear that different frequency oscillations may play the crucial role in the integration of different kinds of brain activity [8, 9]. Hence, individual-specific time-varying patterns of the distribution of spectral power across the cortex may relate to the within- and between-subject differences in functional activity. General understanding in EEG research is that cortical distribution of spectral power varies as a function of time, frequency, state, and experimental condition. There are findings, however, which show that individual-specific patterns of cortical spectral power distribution could be amazingly stable, at least in some experimental conditions [10-12]. It has been shown, for example, that a particular topographic distribution of the EEG spectral power along the antero-posterior cortical axis, like a "fingerprint," distinguishes each individual during non-REM sleep. This individual EEG-trait was substantially invariant across six consecutive nights characterized by large experimentally induced changes of sleep architecture. It has been hypothesized that this EEG invariance can be related to individual differences in genetically determined functional brain anatomy, rather than to sleep-dependent mechanisms [11]. Similarly, Tinguely et al. [12] in their study of functional EEG topography in sleep and waking have found that, although EEG spectral power is modulated by state and sleep pressure, basic topographic features appear to be state-independent. Moreover, a high degree of intra-individual correspondence of the spectral power maps was observed across the studied states. Basile et al. [10] have shown that during a simple visual task individual-specific patterns of spectral power cortical distribution are largely the same as they were during the pre-stimulus baseline. In two different experimental datasets Knyazev [13] analyzed the stability and variability of individual patterns of cortical spectral power distribution across time, experimental conditions, and frequency bands. It has been shown that in spite of high short-term variability, individual-specific patterns of cortical spectral power distribution are remarkably stable across frequency bands, long periods of time, and experimental conditions. These patterns are related to state and trait participant’s characteristics. The antero-posterior spectral power gradient (APSPG) emerged as the most prominent feature associated with important personality dimensions. Relatively higher oscillatory activity in the frontal cortical region was related to Behavioral Inhibition (predisposition to anxiety and vigilance, [14]), whereas relatively higher activity at posterior sites was associated with Extraversion (E). It should be noted in parentheses that higher or lower oscillatory activity is defined here as higher or lower spectral power measured by means of conventional spectral decomposition techniques, such as Fast Fourier Transform (FFT) or wavelet transform. This finding is in line with the literature linking higher or lower activity in the prefrontal cortex (PFC) to approach and withdrawal behavioral tendencies. Indeed, although the most known and popular among EEG researchers theory relates personality traits such as Behavioral Activation and Behavioral Inhibition to asymmetrical frontal activity [15], several researchers have emphasized the importance of bilateral (or symmetrical) frontal EEG activity in the context of personality traits [16-18]. Hewig et al. [17, 18] proposed that low resting bilateral frontal EEG activity in the alpha band is associated with high habitual activity in Gray’s [14] Behavioral Activation System (BAS). The BAS is a dopamine-based motivational system closely related to E [19]. Hewig et al. [17, 18] observed significant negative correlations between scores on Carver and White’s [20] BAS scale and resting EEG alpha power at frontal scalp sites and a positive correlation of similar magnitude between BAS sensitivity and EEG alpha power at parietal and parieto-occipital sites. Moreover, several functional magnetic resonance imaging studies have found correlations between E traits and activity in the anterior cingulate cortex (ACC) during a number of cognitive and emotional tasks [21-24]. The ACC is an area of the PFC that receives strong dopaminergic (DA) inputs [25, 26] and whose activity is influenced by genes controlling PFC DA neurotransmission [27]. There is evidence [28] that activity in this region is reflected in frontal midline EEG activity in the theta band (4-8 Hz). Basing on theses evidences, Wacker et al. [29] hypothesized that a difference between EEG alpha activity at frontal sites minus activity at parietal sites would be maximally suited to capture the aspect of brain electrical activity related to habitual BAS activation. They also hypothesized that the difference between frontal midline theta activity and theta activity at posterior sites would be well suited to capture an aspect of EEG theta activity related to brain dopaminergic functioning. They were able to demonstrate that the negative association between E and the frontal (Fz) minus parietal (Pz) theta activity, which was observed in the placebo group, was completely reversed in the group that received the selective dopamine D2 antagonist sulpiride [29]. Similar effect was observed for the low alpha band (8-10 Hz) but it failed to reach significance level. This intriguing finding implies that the fronto- posterior spectral power gradient (specifically in the theta band of frequencies) may reflect trait-like predispositions which depend on the brain dopamine functioning.

It should be noted that Knyazev [13] used collapsed across all frequencies measure of the antero-posterior gradient (APSPG). In light of the above discussed findings by Wacker et al. [29], a question arises as to whether APSPG in different frequencies would show similar correlations with personality, or (as it appeared in the Wacker et al.’s study) most substantial correlations would be found for the theta band. The latter result would be meaningful given theoretical considerations and empirical evidence linking theta oscillations with emotion processing [30, 31]. Another question is whether the APSPG relates specifically to E (and, more precisely, to agentic E [29]), or is associated with a broader set of personality traits? Depue and Collins [19] note that E can be subdivided into sociability on the one hand and agency on the other. It is the agency facet of E (i.e., a motivational disposition that comprises social dominance, enthusiasm, energy, assertiveness, ambitiousness, and achievement striving), which they regard as the central characteristic of E and for which they propose a dopaminergic basis in their theory. Wacker et al. [29] show that the difference between frontal midline theta activity and theta activity at posterior sites relates specifically to agentic E (which was measured by the Activity scale of the Zuckerman-Kuhlman Personality Questionnaire [32]). On the other hand, Knyazev [13] has found correlations of the APSPG not only with E, but also with facets of Neuroticism and Behavioral Inhibition. The third question is whether individual differences in APSPG (and its association with personality) are stable across different experimental conditions? Knyazev [13] show moderate to strong between-condition correlations in cortical spectral power distribution, but this does not actually mean this distribution does not change in different states. Wacker et al. [29] show that both the magnitude of the frontal minus parietal theta and alpha activity and the strength of its association with E are moderated by different experimental conditions and by administration of the selective dopamine D2 antagonist. Finally, we aimed to explore the cortical/subcortical sources of individual differences in APSPG. Basing on existing literature, we mainly expected to find these sources in the cingulate gyrus. To address these questions, previously collected data [33] were reanalyzed.

MATERIALS AND METHODS

RESULTS

Estimated marginal grand mean (Std. Error) for the percent of trials with a positive APSPG (PAPSPG) value was 44.2 (2.4). Repeated measures ANOVA with experimental condition and frequency band as within-subject factors and PAPSPG value as the outcome showed significant main effects for condition (F = 4.36, df = 1.79, p = 0.020), frequency (F = 23.24, df = 1.76, p < 0.001), and their interaction (F = 13.87, df = 5.51, p < 0.001). PAPSPG was higher for delta and theta (estimated marginal means = 53.4 and 52.9, respectively) than for alpha and beta (estimated marginal means = 35.3 and 35.2, respectively). Due to significant interaction condition x band, the effect of condition was further analyzed separately for each frequency band. It was significant for delta (F = 3.01, df = 2.42, p = 0.044), theta (F = 3.97, df = 2.37, p = 0.017), and beta (F = 19.34, df = 1.82, p < 0.001), but not for alpha (F = 1.09, df = 2.16, p = 0.343) band. In all frequency bands PAPSPG tended to be higher during baseline and ITI than during PSP.

Series of repeated measures ANOVAs were further conducted for each frequency band separately with experimental condition as within-subject factor, gender and psychometric variables as between-subject factors, and PAPSPG as the outcome. In all frequency bands females showed higher mean PAPSPG than males, but this effect was significant only for delta band (F = 9.42, df = 1, p = 0.004). The main effect of E was significant for theta (F = 4.8, df = 1, p = 0.036) and alpha (F = 4.12, df = 1, p = 0.050) bands. The interaction E x condition was significant for delta (F = 3.15, df = 2.63, p = 0.035), theta (F = 3.69, df = 2.43, p = 0.022), and alpha (F = 3.17, df = 2.25, p = 0.042) bands. In all cases extraverts showed lower PAPSPG than introverts, but this difference was most pronounced during the baseline condition. In all other conditions, relative to baseline, PAPSPG tended to decrease in all subjects, but this decrease was sharper in introverts than in extraverts. As a result, the main effect of condition was significant in introverts (F = 3.62, df = 2.05, p = 0.036), but it was not significant in extraverts (F = 1.41, df = 1.76, p = 0.260). When effects of the three E facets were analyzed, only Sociability showed significant main effects for delta (F = 5.83, df = 1, p = 0.022), theta (F = 6.41, df = 1, p = 0.016), and alpha (F = 5.35, df = 1, p = 0.027) bands Fig. (1, 2), whereas effects of Activity and Assertiveness were nonsignificant.

Assertiveness showed significant interactions with condition for theta (F = 3.07, df = 2.38, p = 0.043) and alpha (F = 5.93, df = 2.42, p = 0.002) bands. High as compared with low Assertiveness subjects showed lower PAPSPG during baseline recording, but higher PAPSPG during all other conditions. Main effect of Trait Anxiety was significant only for theta band (F = 6.22, df = 1, p = 0.018). Main effect of BIS was significant for delta (F = 5.48, df = 1, p = 0.025), theta (F = 9.02, df = 1, p = 0.005), alpha (F = 10.02, df = 1, p = 0.003), and beta (F = 5.6, df = 1, p = 0.024) bands. In all cases anxious and high BIS subjects showed higher PAPSPG Figs. (3, 4). There were no significant effects of BAS and Psychoticism.

Finally, sLORETA analyses were conducted to compare cortical sources of individual differences in APSPG values with cortical sources of individual differences on relevant psychometric variables. Baseline APSPG values were averaged across the four frequency bands and used in a regression analysis as the independent variable to predict baseline, ITI, and PSP delta, theta, alpha, and beta current source density estimates. As there were only minor differences in localization of sources for baseline, ITI, and PSP conditions, only results for the baseline condition are reported Fig. (5).

For delta band, maximal correlation (0.31) was found in the cingulate gyrus, BA23 [(X, Y, Z) = (-5, -25, 30)]; for theta band maximal correlation (0.26) was found in the insula, BA13 [(X, Y, Z) = (-35, -20, 15)]; for alpha band maximal correlation (0.38) was found in the anterior cingulate, BA25 [(X, Y, Z) = (0, 10, -5)]; for beta band maximal correlation (0.41) was found in the anterior cingulate, BA24 [(X, Y, Z) = (5, 25, -5)].

The same analysis was conducted using relevant psychometric variable instead of APSPG. Because BIS and Sociability showed strongest associations with APSPG, their combination was used in this analysis (i.e., BIS z-scores - Sociability z-scores). Fig. (6) shows results for the baseline condition.

For delta band maximal correlation (0.58) was found in the posterior cingulate, BA29 [(X, Y, Z) = (-5, -50, 10)]; for theta band maximal correlation (0.64) was found in the anterior cingulate, BA32 [(X, Y, Z) = (10, 35, -5)]; for alpha band maximal correlation (0.67) was found in the posterior cingulate, BA30 [(X, Y, Z) = (-5, -55, 15)]; for beta band maximal correlation (0.51) was found in the anterior cingulate, BA24 [(X, Y, Z) = (5, 30, 0)].

DISCUSSION

In our analyses, we deliberately avoided averaging across trials. Instead we chose a measure of relative frequency of appearance of a particular instantaneous pattern of cortical EEG spectral power distribution which is characterized by relatively higher oscillatory activity in frontal than in posterior cortical sites. Overall, this pattern appeared in approximately half of trials, but there were significant between- and within-subject differences which may help to deduce possible physiological (psychological) concomitants of this pattern. In the within-subject domain, relative prevalence of oscillatory activity in frontal regions was more frequently observed in “passive” states (baseline and ITI) than in states associated with the perception of visual information. This is to be expected given that visual perception is associated with occipital cortical regions. This fact further confirms that brain oscillations have functional significance and do not reflect an “idling” condition.

Effects of personality variables were not restricted to the theta band of frequencies. Similar effects of E were observed in delta, theta, and alpha bands. Effects of BIS were significant in all four frequency bands. It should be emphasized that this similarity refers only to the pattern of individual differences in the antero-posterior spectral power gradient values and its correlations with personality variables. It does not mean that within each individual subject the actual antero-posterior distribution of spectral power was identical in different frequencies. It is common knowledge that in healthy subjects during rest alpha power is maximal in posterior, whereas delta and theta power is higher in the frontal cortex areas.

The observed uniformity of relations between personality and the antero-posterior spectral power gradient in different frequencies seems puzzling given vast literature highlighting different functional correlates of different frequency oscillations. It appears that individual-specific pattern of cortical distribution of oscillatory activity and reciprocal relationships between different frequency bands are two different dimensions. Whereas the former reflects general level of oscillatory activity in different cortical regions, the latter relates to specific between-band relationships which could be observed within each region. It seems that the magnitude of the overall (across all frequencies) oscillatory activity is an important characteristic, which probably reflects intensity of functional involvement of a particular cortical region.

In our study, APSPG was most consistently associated with Sociability and BIS. Specifically, relatively higher oscillatory activity in frontal than in posterior cortical sites was more frequently observed in individuals with higher behavioral inhibition and lower sociability than in disinhibited and sociable individuals. This pattern of results somewhat differs from Wacker et al.’s [29] suggestion that APSPG should be most closely associated with agentic aspects of E. In our study, these aspects were measured by Activity and Assertiveness scales. Activity did not show significant associations with PAPSPG, whereas Assertiveness showed the expected association only during baseline condition.

Analysis of cortical sources of inter-individual differences in APSPG values showed that for all but theta bands these differences were associated with differential activity within the cingulate cortex. For theta, cortical sources were found in the insula. Amazingly similar (except for theta) results were obtained when APSPG was replaced with the relevant psychometric variable. The strength of associations with cortical sources current density was higher for the psychometric variable than for APSPG. Thus, maximal correlations for APSPG were 0.31, 0.26, 0.38, and 0.41 for delta, theta, alpha, and beta bands, respectively, whereas for the psychometric variable they were 0.58, 0.64, 0.67, and 0.51. It should be noted that given low spatial resolution of sLORETA and the known caveats inherent in inferring brain electrical sources from scalp EEG data, precise localization of maximal correlations should be interpreted with caution, and the most robust result seems to be that higher behavioral inhibition and lower sociability are associated with higher activity in different parts of the cingulate cortex. The cingulate gyrus is involved in various functions, such as attention, pain, response selection, maternal behavior, vocalization, sensorimotor function and autonomic control [43]. However, it is generally acknowledged that different parts of the cingulate gyrus play a prominent role in conscious and unconscious perception of emotion. In 1937, James Papez [44] proposed that the dorsal anterior cingulate cortex was the seat of emotional experience. Although current models of how the brain mediates emotion are very different from the Papez model, this hypothesis was surprisingly prescient [45]. Cingulate lesions result in profound behavioral disturbance with abnormal expression and experience of affect [46]. The posterior cingulate is traditionally considered a part of the limbic system and has abundant connections with the hippocampus and the amygdala. The anterior cingulate plays an important role in arousal, attention, motivation, emotion, and feeling [45, 47, 48], in the interaction between emotion and attention [49, 50], in the top-down modulation of intense emotional responses [51, 52]. It has been suggested that the anterior cingulate has a superordinate role in executive control of attention [45]. According to this view, emotion, pain, or other salient exteroceptive or interoceptive stimuli provide moment-to-moment guidance regarding the most suitable allocation of attentional resources [45]. This idea is consistent with Paus’ [53] view that the dorsal anterior cingulate is fundamentally involved in translating intentions into actions in the context of emotional arousal. Being a place of convergence for motor control, emotion and cognition [54], the anterior cingulate may participate in emotional response inhibition [55] and in inhibitory control of movement [56]. Subjects with low sociability and high behavioral inhibition may be particularly vigilant and experience increased emotional tension in the environment of a psychophysiological laboratory which may show up in increased oscillatory activity within the cingulate cortex. Noteworthy, the difference between high and low Sociability subjects is most pronounced during baseline recording when the subjects’ attention is not captured by an external task; this difference substantially diminishes when attentional resources are directed to the task. As for the association between APSPG and cortical theta sources in the insula, this result should be treated with caution until it is replicated in some other study, but it generally corroborates the view that APSPG-related individual differences relate to differences in emotion processing because insula is a higher association area for bodily signals [57] and participates in the neural basis of background feelings [45, 58].

CONCLUSIONS

In conclusion, this study analysis shows that although cortical distribution of spectral power is relatively stable across time and experimental conditions (Knyazev, 2009), the APSPG is generally more pronounced during resting baseline than during ITI and PSP. Its association with personality variables is similar for different frequency bands (see Figs. 2 and 4) and is mostly preserved during different experimental conditions (see Figs. 1 and 2). Source localization analysis shows that both high BIS/low Sociability and high APSPG are associated with higher oscillatory activity in medial cortices associated with emotion processing (mostly cingulate gyrus). This association could be tentatively explained by higher vigilance and emotional tension in introverted and behaviorally inhibited subjects, although this issue clearly needs further exploration in different experimental conditions and with different psychometric instruments.

REFERENCES