Abstract

Oscillations and temporal structure in the brain

According to the traditional view, brain function is primarily described on the basis of functional anatomy. Anatomy and functional connectivity can be considered the spatial or geometrical dimension of the mind. However, for a more comprehensive understanding, an additional dimension must be considered: time. The brain generates its own temporal structure, which is largely organized by oscillations (e.g., [1]). This aspect has become an important issue in the investigation of perceptual and cognitive processes (for recent reviews see, for example, [2,3]). In this review, the focus is on alpha-band oscillations as the dominant oscillations in the human brain, with a mean frequency of approximately 10 Hz.

In the sections that follow, I will argue that alpha-band oscillations reflect the temporal structure of one of the most basic cognitive processes, which may be described as ‘knowledge-based consciousness’ and which enables ‘semantic orientation’ via controlled access to information stored in a complex knowledge system (KS; see Glossary).

Alpha-band oscillations and two types of event-related amplitude responses

Alpha-band activity is particularly interesting because it is the only frequency domain (with the exception of slow beta, a frequency higher than alpha, up to approximately 20 Hz) that responds to a stimulus and/or task demands either with a decrease or increase in amplitude/power, termed event-related desynchronization and synchronization, or ERD and ERS, respectively. The response of other frequencies (particularly in the delta, theta and gamma frequency range) is ‘simply’ characterized by ERS (see [4,5]). This is interesting because one implicit assumption in EEG research is that oscillations have an impact on information processing that is proportional to their magnitude. In this sense, ERS, but not ERD, would have to be considered the ‘active’ task-relevant alpha-band response. This, however, is in contradiction with the classical view (e.g., [6]) which holds that suppression of alpha-band activity (ERD) is the ‘typical’ event-related alpha response. This view comes from the basic observation that alpha amplitudes are large when the eyes are closed and become suppressed when eyes are opened. It has been known since the early days of EEG research that this type of alpha suppression is not due to visual stimulation because it can also be observed in a completely darkened room [7,8]. ERD can be observed in response to a variety of different tasks, and it was assumed that it reflects cortical activation, or more precisely, cortical excitation. More recently, however, it was recognized that under certain task demands ERS (instead of ERD) can be observed.

Alpha phase and the inhibition timing hypothesis

ERS and other measures of power are insensitive to phase. However, it is the very nature of an oscillation to reflect rhythmic changes. With respect to alpha oscillations, as inhibitory oscillations, this means that the change is between minimal and maximal inhibition (see [19] and Figure 1).

Open in a separate window

Figure 1

The inhibition timing hypothesis. (a) Illustration of the main assumption of the inhibition timing hypothesis. Three cases are distinguished. Case 1: In task-irrelevant networks, the amplitude of alpha oscillations is small and has no impact on cell firing rate, which is symbolized by short vertical lines representing action potentials (APs) in the lower panel. Case 2: In task-relevant networks, the amplitude of alpha oscillations increases and starts to inhibit the generation of APs in target cells during the inhibitory phase of the oscillation. Depending on the excitation level of target cells, the impact of the inhibitory oscillation is different (cf. Cell 1 and 2). With an increase in amplitudes, the inhibitory baseline increases, which means that inhibition increases not only during the phases with maximal inhibition (plotted here as troughs), but also during those with minimal inhibition (plotted here as peaks). Case 3: If inhibition increases further, all target cells are silenced. This case is assumed for neuronal structures that are potentially competing relative to the processing of a task. A good example is the inhibition of potentially competing regions in the ipsilateral hemisphere. Adapted, with permission, from [30]. (b) An alternative interpretation of Case 3. The increased alpha amplitudes during a state of inhibition may stem from spatial summation (at the level of scalp electrodes) of phase-coherent alpha sources that lead to a monotonous firing pattern in a population of target cells. If there is no variation in the AP firing pattern, no information can be transmitted. Such a case may emerge in task-irrelevant structures in which the excitation level of a population of principal cells is low and does not vary systematically.

The two faces of inhibition: selective activation vs blocking of information processing

In an ideal situation, excitation and inhibition may influence each other in such a way that inhibition is silencing weakly excited cells, but operates to induce a pulsed pattern of action potentials (APs) in cells with a higher excitation level, as is illustrated in Case 2b of Figure 1a. Thus, the time and direction of a change in inhibition – described by phase – is functionally related to the timing of neuronal activation processes. This interpretation can nicely explain the observation that inhibition is an active process for information processing that helps to increase the signal to noise ratio (SNR). In this sense, inhibition and excitation depend on and are ‘tuned’ by each other (e.g., [20,21]) to increase SNR.

It is important to note that this ideal situation cannot simply be explained by an increase in amplitude. Consider Case 1 and Case 2a in Figure 1a. An increase in amplitude induces ‘timing’, which is apparent in the pulsed AP pattern for Cell 1 and Cell 2. However, a further increase in amplitude would not lead to inhibition (in the sense of silencing less excitatory cells), but to even more precise timing. A hypothetical assumption to overcome this problem is to assume a baseline shift as is illustrated in Case 2b and 3 of Figure 1a, which means that the phase with minimal inhibition also exhibits an increase in the magnitude of inhibition. This could be a mechanism that enables the ‘silencing’ of those principal cells which have a comparatively low level of excitation (see Cell 1 in Case 2b).

For task irrelevant/competing structures yet another mechanism might be considered. The excitation level of principal cells will generally be low and an increase in amplitude will lead to precise timing, but without any systematic variation in the temporal AP pattern. The reason is that the excitation level of principal cells is not modulated (as they are – by definition – not involved in processing a task) and, thus, the AP firing pattern will not differ systematically between cells. Because information is most likely encoded by variations in interspike intervals and the number of spikes in bursts, a completely monotonous AP pattern as illustrated in Figure 1b will not allow information to be transmitted. Such a scenario might be consistent with the suggestion that oscillations play a role for phase coding, as suggested by Nadasdy [22] for fast frequencies in the gamma range. The central idea is that topographical phase differences in traveling waves code information. A stationary wave, characterized by a lack of topographical phase differences, will not be able to code information but may lead – via spatial summation – to a large amplitude at a scalp electrode. A lack of phase coding may be considered a state that blocks the processing of information. The finding that alpha hypersynchronization is associated with a loss of consciousness [23] is well in line with these considerations.

The physiology of brain oscillations is very well investigated, particularly in animal research, for certain types of oscillations, especially for theta, gamma, and sleep spindles (for reviews see, for example, [24,25]). Only a few studies have focused on alpha oscillations (Box 1). Although there is evidence that alpha-band oscillations have an inhibitory influence on the generation of action potentials as was demonstrated by Haegens [26] (Figure 2), the exact physiological mechanisms that generate alpha-band activity are not yet known.

Open in a separate window

Figure 2

A recent animal study by Haegens et al.[26] demonstrates nicely that alpha activity drives spiking and that the power of alpha oscillations is associated with the rate of spiking. Monkeys were trained to perform a vibrotactile discrimination task between two stimuli S1 and S2. Local field potentials and spikes were recorded simultaneously. (a) Spike field coherence in M1 (contralateral to the responding hand) is most pronounced in the alpha frequency range. This indicates that alpha-band activity drives action potentials. The black line represents the spiking rate during task performance. (b) Power spectra for a baseline, the retention period, and the decision period exhibit a clear alpha peak at approximately 11 Hz. During the decision period, a harmonic peak at approximately 22 Hz emerges, probably reflecting the preparation of the motor response. Note that spiking rate is negatively associated with alpha power. Large alpha power during retention is associated with reduced spiking, whereas a decrease in alpha power during decision is associated with enhanced spiking. Reproduced, with permission, from [26].

Box 1

The physiological basis of alpha-band oscillations in animal research

Early research [101] tried to relate alpha-band activity to spindle activity and to thalamo-cortico-thalamic re-entrant loops. Nevertheless, it is clear that alpha-band oscillations and spindle oscillations have a strikingly different physiological basis [102]. Alpha-band activity certainly plays an important role in thalamo-cortical structures [103–107]. Thalamic nuclei, particularly the lateral geniculate, are known to play an important role in driving alpha activity [108]. Recent findings suggest that the thalamic pulvinar nucleus modulates alpha synchrony between cortical areas as a function of attentional demands [109]. Nonetheless, the functional meaning of alpha-band oscillations appears closely associated with activity in cortico-cortical networks [110–112]. There is promising recent research about the cortical sources of alpha-band activity, showing that it is present at different cortical layers and is capable of driving multi-unit activity in a task dependent manner [26,112]. Most importantly, Haegens et al. [26] were also able to demonstrate that during periods of increased alpha activity, multi-unit activity decreases (see Figure 2 in main text). This is probably the first physiological evidence for an inhibitory role of alpha-band activity. Up to the present, there is only indirect evidence that alpha-band oscillations may stem from rhythmic GABAergic activity. For example, Lörincz et al. [105] demonstrated that a subtype of excitatory thalamo-cortical neurons burst in alpha frequency and are driving inhibitory interneurons.

The cognitive significance of alpha-band oscillations

When trying to evaluate the possible cognitive significance of alpha-band oscillations the following two questions should be considered. First, is the cognitive role of alpha oscillations related to a well described cognitive ‘domain’, such as perception, attention, working memory (WM), and long-term memory (LTM), or is it related to functions that are yet not well described or difficult to describe? I will argue for the latter. Second, in light of the fact that the inhibitory function of alpha oscillations appears well supported and elaborated [27], the question which cognitive processes are best compatible with the inhibition timing function must be asked.

The first question is complex and difficult to answer. The reason is that alpha oscillations appear to respond to almost all of the aforementioned cognitive domains. I will argue that alpha-band activity reflects a special class of processes – related to the controlled access to and retrieval from the KS – which are yet not well described. These processes are very basic and, thus, underlie many other cognitive processes. Let us consider the traditional alpha response, ERD, as an example. ERD can be observed in response to a large variety of stimuli, but also to cognitive demands without stimulation (e.g., [28,29]). A first hint to a possibly specific function of alpha-band activity is the very reliable finding that during retrieval from LTM the magnitude of ERD varies as a function of the semantic content of the information that is retrieved. The more semantically integrated the retrieved information is, the larger the ERD magnitude is (at approximately 200 – 500 ms poststimulus; for reviews see [5,19,30,31]). Because ERD can be interpreted as release from inhibition, this finding simply means that retrieval of semantically well integrated information elicits more cortical excitation than less integrated information. This finding is interesting because of its cognitive selectivity.

The second question is easier to answer. If alpha-band activity is indeed inhibitory, then it should underlie a cognitive process that reflects the two aspects of inhibition, timing and blocking information processing as described above. One of the most promising candidates is attention, because one of its central functions is to enhance and ‘sharpen’ processes within the attentional focus and to block processing outside the focus. When focused on external information, attention is usually under top-down control and operates as an ‘interface’ with the external world. In this way, attention interacts closely with perception and is subject to capacity limits of the ‘attentional buffer’ (which may be considered a ‘sub-buffer’ of iconic memory, closely related to the KS, but not to WM). In the internal and external processing mode, attention is anticipatory with a focus on the type of information that is conveyed by a stimulus and the time point of its appearance. These two aspects of attentional selection are referred to as anticipatory and temporal attention and are suggested to be closely associated with alpha-band oscillations.

For a long time the alpha-band frequency range has been associated with attention [7]. However, attention is a very complex concept comprising very different functions. Here, I will argue that the two fundamental functions of attention as filter (suppression and selection) enable selective access to the KS and operate according to the proposed inhibition timing function of alpha-band activity [19,30]. Brain processes related to the enhancing function within the attentional focus are characterized by a phase response (reflecting the ‘timing’ aspect of alpha-band oscillations; see also [32]) and a subsequent ERD. On the other hand, brain processes related to the suppressing function of attention are primarily associated with an increase in alpha amplitude (ERS), reflecting the inhibitory aspect of alpha-band oscillations (see also [32,33]). A good example for this antagonistic ERD/ERS response is the concept of focal ERD/surround ERS, suggested by Suffczynski et al. [34] to emphasize that cortical activation (reflected by ERD) may be more focused when surrounded by inhibition (reflected by ERS).

Alpha-band activity is associated with certain types of attention only

In support of the notion that alpha-band activity reflects a specific aspect of attention – and not attention in general – is the fact that monitoring new information reliably elicits a strong task- and event-related increase in the theta frequency band (approximately 6 Hz). A good example is the frontal midline theta-band activity that can be observed in response to increased WM demands (e.g., [35] and Figure 4b) reflecting sustained attention to the processing of new information. It should be mentioned that in these tasks, alpha-band activity responds with ERS, most likely reflecting the suppression of task-irrelevant structures and not the enhancing aspect of attentional selection.

Open in a separate window

Figure 4

The harmonic/golden mean frequency architecture around the alpha frequency range. (a) Traditional frequency bands can be ‘reconstructed’ on the basis of three assumptions. (i) The alpha band has a mean center frequency of approximately 10 Hz. (ii) The center frequencies of all neighboring frequency bands are harmonics of the alpha frequency band (see red vertical bars). (iii) The bandwidth is calculated on the basis of golden mean (g = 1.618) frequency relationships to ensure minimal interference between bands. For example, 10 x g ≈ 16. 2 Hz is a frequency maximally separated from the alpha band and marks the lower end of the beta band. The golden mean relationships are indicated by blue bars. (b) Harmonic relationships are frequently observed. An interesting example is the appearance of frontal midline theta-band activity at approximately 6 Hz and posterior alpha-band activity at approximately 12 Hz during the retention period of a demanding working memory task. Adapted, with permission, from [35]. Another example of a harmonic relationship with alpha is shown in Figure 2b.

Temporal attention and alpha phase

The relationship between alpha-band activity and attention is not restricted to task-related changes in amplitude; it also comprises phase. As the findings in Hanslmayr et al. [38] suggest, focused attention elicits large phase locking during the processing of a target stimulus without a concomitant amplitude change. This phase response may be interpreted to reflect temporal attention. Thus, alpha phase should also play a crucial role for the attentional blink (AB) phenomenon, which is one of the most important experimental examples of temporal attention. The AB represents reduced ability to report a second target (T2) after identifying a first target (T1) in a rapid serial visual presentation (RSVP; Figure 3a) of stimuli. The explanations of the phenomenon proposed so far have focused primarily on cognitive aspects, such as attentional filter, capacity limitation, and retrieval failure processes (for reviews see [42,43]). More recently, the role of brain oscillations has also started to be considered (for a review see [44]).

Open in a separate window

Figure 3

The influence of temporal expectation on alpha phase. (a) Single trial structure of an attentional blink (AB) task. The rapid serial visual presentation (RSVP) of stimuli is typically at a rate of approximately 10 items per second. Two targets, T1 and T2, have to be reported after the RSVP. An AB trial is characterized by failure to report T2 after T1 has been detected. (b) For a mean alpha frequency of 10 Hz, phase at stimulus onset around the positive peak (coded as 0°) represents a favorable phase. The idea is that, in this case, alpha-band activity will also exhibit a positive peak at 100 ms post-stimulus that coincides with the appearance of the P1. (c) The rapid presentation rate leads to an occurrence of a steady state visual evoked potential (SSVEP). The evoked potential shown here is filtered in the broad alpha range (8–12 Hz). Broken vertical lines indicate stimulus onset. Note the appearance of a pronounced SSVEP with phase entrainment at the negative alpha peak (approximately 180°) after the appearance of T1. The horizontal red and green lines below the SSVEP indicate time windows of significant phase locking. Note that AB and NoAB trials differ around the appearance of both targets, T1 and T2. (d) and (e) Mean phase angle for the 8-12 Hz filtered data. AB single trials showed significant deviation from uniform distribution, with a mean phase angle of 245° for T1 (d) and 161° for T2 (e) onset. NoAB trials were uniformly distributed (right column). Parts of the Figure are adapted, with permission, from [54].

The central issue here is that the AB depends on a stimulation frequency of approximately 10 items per second, which equals a mean alpha frequency of approximately 10 Hz. The interesting point here is that a fast visual stimulation frequency changes the event-related potential in the sense that the appearance of the typical P1/N1 waveform (consisting of a positive wave at approximately 100 ms and a negative wave at approximately 160 ms post-stimulus; [45]) disappears and a steady state visual evoked potential (SSVEP) emerges (e.g., [46,47]). The properties of driven EEG activity are well documented (e.g., [46,48,49]; for a recent review see [50]) and show in general that the alpha frequency range is the most responsive. The hypothesis here is that entrainment of alpha phase at a certain phase angle, which is associated with the inhibitory phase, is responsible for the AB (Figure 3b). This assumption rests on findings suggesting that the phase angle of alpha-band activity at stimulus onset predicts target detection performance. As an example, Mathewson et al. [51] found that, for undetected trials, the phase at stimulus onset was different from that of detected trials (for similar results, see [52]). When the target could not be detected, a prominent negative peak at stimulus onset was associated with significantly reduced P1 amplitude. As illustrated in Figure 3b, this may suggest that, in trials where the phase of alpha-band activity at stimulus onset interferes with the generation of the P1, the stimulus will not be detected. The crucial consideration then is that ongoing alpha-band activity at 10 Hz (and a period of 100 ms) exhibiting a negative peak at stimulus onset will develop a negative peak at 100 ms post-stimulus. This would lead to suppression of P1 amplitude and failure to detect the target stimulus, because there is evidence that the P1 reflects early processes of stimulus categorization (e.g., [30,53]). A recent study by Zauner et al. [54] tested whether failure to detect T2 (in AB trials) is associated with alpha phase-locking and an entrainment at a phase angle that approximates the negative alpha peak. The findings clearly supported this hypothesis (Figure 3c-e). In the time window around T1 and T2, significant phase-locking (as measured by the phase locking index, PLI) was observed for AB trials only. Most importantly, the phase angles during the onset of T1 (detected) and T2 (undetected) stimuli were different. For T1 stimuli, mean phase angle was at 245°, which is beyond the positive peak in the negative-going slope. For undetected T2 items, the mean phase angle of 161° is close to the ‘unfavorable’ negative peak of 180°. Thus, the critical difference between T1 and T2 is a combination of phase-locking and phase angle (Figure 3). In general, these findings are well in line with the notion that perception is discrete and not continuous (e.g., [55,56]).

During the AB task, the attention-related alpha response is restricted to phase and does not comprise amplitude. There are, however, also changes in amplitude – preceding the RSVP – that are associated with AB magnitude (ABm). For example, McLean et al. [57] observed a complex interaction between alpha ERD (measured in a foreperiod of 2 sec before the RSVP started), ABm, and lag [57]. At short lags (351 ms after T1), a large ERD was associated with low T2 accuracy (high ABm), but at long lags (936 ms after T1), a large ERD was associated with high T2 accuracy. The authors interpreted this finding in terms of an attentional over-investment [58,59] that leads to an increase in ERD in a foreperiod preceding the RSVP, which in turn leads to an increased ABm at short lags. In addition, they observed that resting alpha power is positively associated with ABm [60].

Alpha-band oscillations, semantic orientation, access to the knowledge system (KS), and the P1

Access to the KS may be an event-related or a continuous process. In the latter sense, the alpha-band frequency range reflects continuous ‘semantic orientation’, which represents the ability to be consciously oriented in time, space, and with respect to the meaning of all entities surrounding the individual (note the similarity with the proposed function of a default mode system [67]). This ability requires a certain kind of attention (see above) and is ‘knowledge based’, which means that semantic orientation provides individuals with the ability to selectively access stored information that represents the meaning of sensory information and ‘higher order information’, such as language, mathematics, and geography.

It is interesting to note that the suggested association between semantic information as integrated knowledge [68] and a knowledge based consciousness plays a key role in Tononi's integrated information theory of consciousness [69,70]. The central idea is that ‘consciousness is integrated knowledge and that its quality is given by informational relationships [71].

Selective access to the KS is thought to depend largely on inhibiting task-irrelevant memory entries. Thus, periods of prolonged access should be associated with ERS, reflecting increased alpha-band activity. A nice anecdote to illustrate this idea is a report by Penfield and Jasper ([8], p. 189) about EEG recordings from Albert Einstein. Einstein produced continuous alpha waves during the solution of complex mathematical tasks which he was skilled and trained to solve. When his alpha-band activity suddenly disappeared, he was asked what had happened because he appeared nervous. He replied that he had just become aware of a mistake in the calculations he had recently done. Another nice example that can be observed under controlled experimental conditions is the increase in alpha-band activity during divergent thinking which requires top down-controlled [72] and prolonged access attempts to long-term memories (see [28,29]).

Event-related access to the KS can be easily studied by using object recognition paradigms. The early event-related EEG response reflects sensory processes and it is not obvious when access to memory begins. One empirical approach to de-confound memory access from sensory processes is to present distorted objects, starting with a high level of distortion and ending with the undistorted presentation of the object. This approach was used in a study by Freunberger et al. [41]. On each trial, four items (representing the same object at decreasing levels of distortion) were presented one after the other. The interesting finding was that the first item – which could never be recognized – elicited neither a P1 nor ERD. Thus, neither the P1 nor ERD can be considered EEG responses that are stimulus evoked or stimulus induced (for reviews, see [17,30]). The working hypothesis is that the P1 reflects early stimulus categorization and the onset of ERD retrieval.

Alpha-band oscillations and the frequency architecture of the brain

Some researchers have criticized the traditional approach of investigating strictly defined frequency bands while ignoring the functional interplay between frequencies [25]. This critical view is well justified, because there is hardly any doubt that different frequencies interact in some way and represent different aspects of brain processes. However, so far, there is no global brain theory in sight. Nonetheless, such an approach may be possible if one considers fundamental aspects of between-frequency interactions and the possibility that a certain frequency architecture enables a balanced coupling and de-coupling between oscillations.

Research on brain oscillations suggests that different frequency domains are associated with different classes of cognitive processes. As an example, theta oscillations appear to be related to the processing of new (episodic) information, whereas alpha-band activity is associated with access to information that represents knowledge of the environment (for reviews, see, for example, [5,30,31,76]). Beta-band activity (approximately 20 Hz) may reflect the cognitive control of motor activity and gamma (approximately 40 Hz and higher frequencies) is probably associated with a variety of perceptual and cognitive processes in more localized cortical networks. It is important to note that frequency is related to the size of neural networks in a way that smaller networks oscillate with a higher frequency than larger networks [1,77–80]. A crucial implication is that different frequency domains interact in terms of cross-frequency (CF) coupling [80]. The problem, however, is that spurious (‘unwanted’, ‘erroneous’) CF synchronization may cause considerable interference between oscillatory processing domains. Depending on their numerical ratio, the excitatory phases of two frequencies, f1 and f2, will either meet frequently and regularly, as is the case for harmonic frequency relationships, or infrequently and irregularly, as is the case for non-harmonic relationships. Harmonic coupling provides an optimal basis for a functional interchange between two oscillatory systems and is frequently observed (e.g., [81,82] and the examples shown in Figure 2b and 4b). Most importantly, Palva and Palva have demonstrated that phase-coupling between alpha-band (at approximately 10 Hz) and beta-band (at approximately 20 Hz) oscillations changes in a task dependent manner ([83,84]; for reviews, see [71,85]). Non-harmonic coupling blocks an interchange only to some extent because of spurious CF synchronization. Pletzer et al. [86] demonstrated that, in a mathematical sense, two frequencies f1 and f2 will never synchronize if the frequency relationship for f1, f2 (f2 > f1) equals the golden mean (g = 1.618) and that there is no other frequency relationship that is better capable of avoiding spurious CF synchronization (see also [87], for empirical findings).

The interesting point is that the traditional EEG frequency bands can be explained by a frequency architecture that is centered around alpha-band activity (Box 2 and Figure 4). The assumption is that the alpha frequency domain, as a basic process of the conscious brain, interacts best with those frequency domains that have a harmonic relationship with the alpha domain and that the separation between frequency domains is obtained with frequency ratios equaling the golden mean. It is a consequence of this assumption to expect that the frequency architecture changes when consciousness changes. For example, in slow wave sleep a frequency structure is expected that is not centered around alpha-band activity (Box 3).

Concluding remarks and suggestions for future research

Controlled access to the KS, operating under anticipatory and temporal attention, represents the class of cognitive processes that are reflected by alpha-band activity. If this is a valid conclusion, future research on the alpha frequency domain will shed new light on a cognitive domain that has not been well described yet. Access is associated with an inhibitory filter that is reflected in synchronized alpha activity (a phase response and/or ERS) that is followed by ERD (approximately 250 to 400 ms post-stimulus). In traditional research, ERD was considered the ‘typical’ alpha response. The interpretation suggested here is that ERD reflects release from inhibition and, thus, is only part of a more complex alpha process that also comprises inhibition and timing that precedes ERD.

Access to the KS is not only an event-related, but also a continuous process. For adult humans, it is a basic and also largely automatic process that may in everyday language be described as ‘knowing’. One probably could also use the term ‘knowledge-based consciousness’ and contrast it to a ‘WM- or emotion-based consciousness’. Over the life span, alpha activity becomes dominant during early adulthood [88–90], probably reflecting the emergence of an established KS. In childhood, slower frequencies in the broad theta frequency range dominate. They probably reflect the dominance of a WM-based consciousness, because these slower frequencies are closely associated with the encoding and retrieval of new information [5]. These considerations may also explain why for some people alpha-band activity is not dominant (e.g., [88]). Possibly, their consciousness is more WM- or emotion- and less knowledge-based. Another implication is that individual differences in alpha-band activity could reflect trait differences related to the (dominant) cognitive processing mode and personality. Trait differences related to ABm and alpha-band activity are a good example (see the section on resting alpha power and the attentional buffer).

The considerations presented here have a variety of implications, which offer promising avenues for future research. For example, because alpha-band activity is associated with access to the KS, the question which processes underlie WM and the encoding of new information arises. The basic idea here is that the KS and WM interact in a way that traces stored in the KS are used for short-term storage (see [91,92]). The implication is that CF coupling between alpha, theta, and gamma oscillations is expected to reflect the interaction between WM and the KS (cf. [15,93]). Another interesting question is which processes reflect spreading activation within the KS. Furthermore, the predicted baseline shift of alpha activity during increases of inhibition in task-relevant networks (Figure 1a) may also be an interesting phenomenon for future research. This shift appears to bear a close resemblance to findings revealing an asymmetric amplitude fluctuation of alpha activity that possibly explains the generation of slow event-related potentials [94,95].

Further implications are related to the hypothesis that early ERP components are generated (at least in part) by oscillations in different frequency domains and to the investigation of traveling waves. All available evidence suggests that the EEG is not stationary. However, neither EEG nor MEG is well suited to study this question, because the folding (gyrification) of the cortex leads to the emergence of radial and tangential dipoles. The first type of dipole underlies the EEG signal, the second the MEG signal. The simultaneous recording of both signals and new software technology make it possible to observe complex traveling waves as they spread over the ‘flattened’ cortex [96] and allow fascinating new insight in the temporal dynamics of the brain.

Acknowledgments

References