Superior Index Go to the next: Chapter 3
Print Files: A4 Size.
This chapter asks the question "can we tell what state of consciousness a person is in just by examining their EEG?" The answer it comes up with is a slightly qualified "yes".
The three major states of consciousness are usually regarded as being waking, sleeping and dreaming. As it turns out, these states, and various substates of them, are characterised by quite clearcut differences in the pattern of electromagnetic oscillations that can be recorded from the scalp, by either EEG (electroencephalograpy) or MEG (magnetoencephalography). Table 2. shows the various frequencies of oscillation that can be detected in EEG records and their proponderance in various states and substates of consciousness. In general, waking and dreaming are characterised by low-amplitude, relatively high frequency oscillations that are often called "desynchronised" EEG. Non-dreaming sleep involves so-called "synchronised" EEG waves, which are larger and slower.
| Table 2: EEG Rhythms | ||
| Name | Oscillation frequency (cycles/sec) | Conscious correlates |
| Fast waves | Over 8 Hz | Waking |
| Alpha | 8 - 13 Hz | Relaxed wakefulness, eyes |
| [Kappa] | [8 - 13 Hz] | closed |
| Mu (Rolandic) | 7 - 11 Hz arch-shaped waves | [Artefact of eye movements] |
| Tau | 8 - 10 Hz | Blocked by body movements |
| Beta | 13 - 30 Hz | Blocked by audition (?) |
| Gamma | 30 - 80 Hz (but often referred to simply as 40 Hz) | Wakefulness. Active cognition, during either wakefulness, REM sleep or occasionally slow wave sleep |
| Lambda | Sharp saw-tooth transients | Scanning visual images: rare |
| Slow waves | Under 8 Hz | Drowsiness and Sleep |
| Theta | 4 - 8 Hz | Drowsiness: alpha dropout |
| Sleep spindles | 1-2s bursts of 7-14 Hz, every 5-10s | Mostly Stage 2 sleep |
| Delta | 1 - 4 Hz | Stage 3 and 4 (deep) sleep: the deeper the sleep, the larger and slower the waves |
Waking and dreaming are obviously different states of consciousness, but they are similar in that subjective sensory and cognitive experiences are a prominent feature of both. Interestingly, the EEG of the two states is indistinguishable, apart from the rapid eye movements that give the dreaming state the label REM sleep. The EEG state which is common to dreaming sleep and waking is often called desynchronised EEG, because it is composed of low amplitude, relatively fast oscillations that have some of the characteristics of 1/f noise. The frequency range of the oscillations is from around 20 to 80 Hz. Such oscillations are now generally termed gamma waves. Sometimes they are referred to under the loose rubric "40Hz", although only some are actually oscillating at exactly 40 Hz.
Two major differences between waking and dreaming are:
As mentioned, dreaming mostly occurs during an EEG state that is associated with rapid eye movements (REM) but is otherwise indistinguishable from waking EEG. This is inferred because people woken out of a state of REM sleep usually report that they have just been dreaming. Because dreams are rarely reported on awakening from other stages of sleep, it used to be thought that dreaming only occurs in REM sleep. However there is now considerable evidence that dreams can occur during all stages of sleep, including deep, slow-wave sleep (Antrobus, 1983 [15]; Cavallero, Cicogna, Natale, Occhionero, & Zito, 1992 [51]; Cavallero, Foulkes, Hollifield, & Terry, 1990 [50]; Foulkes & Schmidt, 1983 [89]). Little work has been done on the EEG concomitants of non-REM (NREM) dreaming, but an attenuated gamma activity can sometimes be seen superimposed on the slow wave EEG (Llinas & Ribary, 1993 [173]). It is tempting to suppose that this corresponds to NREM dreaming.
Perhaps surprisingly considering the recent flurry of hypotheses about the functions of so-called 40-Hz, or gamma waves, the cellular mechanisms generating them are poorly understood. Direct experimental observation shows that thalamic cells which project to the cortex do oscillate at around 40 Hz, owing both to intrinsic mechanisms due to membrane conductances and extrinsic network mechanisms (Steriade, 1999 [278]). For example, cells in the intralaminar nucleus of the thalamus spontaneously discharge 40 Hz bursts of spikes firing at about 1000 Hz, during both waking consciousness and REM sleep (Steriade, Curro Dossi, & Contreras, 1993 [276]). Some of these intralaminar cells project, with abnormally high conduction velocities, to the association cortex (Steriade, Dossi, Pare, & Oakson, 1991 [275]). In this context it is interesting that the intralaminar nucleus of the thalamus seems to be one of the few structures in the brain which is absolutely necessary for the preservation of waking consciousness (Bogen, 1995a [30]; Bogen, 1995b [31]; Bogen, 1997 [32]). This suggests that the intralaminar nucleus of the thalamus might in fact be the pacemaker for cortical 40 Hz oscillations, as the reticular nucleus is for sleep spindles and delta oscillations (see later sections). There is evidence from MEG recordings (Ribary, et al., 1991 [236]) that mass oscillations at around 40 Hz sweep in waves across the cortex from front to back (although this evidence is apparently controversial (Hari & Salmelin, 1997 [120])). These studies show a striking coherence between cortical and thalamic sites of origin, which again suggests that the oscillation is a recurrent thalamocortical one.
However, this reasonably good evidence that thalamic input is important in generation of gamma oscillations notwithstanding, several models for strictly cortical generation of gamma waves have been promulgated. Such models differ mainly in whether they see the major factor in the generation of gamma frequency oscillations as being cortical network properties (i.e. local feedback loops) (Bressler & Freeman, 1980 [35]; Eeckman & Freeman, 1990 [77]; Freeman, 1991b [98]) or the intrinsic oscillatory properties of some cortical cells (Llinas, Grace, & Yarom, 1991 [172]), or both (Jefferys, Traub, & Whittington, 1996 [130]; Whittington, Traub, & Jefferys, 1995 [308]). Each of these purely cortical models is effective in generating gamma oscillations theoretically and it is quite likely that all of them reflect some aspects of the truth, perhaps to differing extents in different regions of the brain. The only mathematical EEG simulation that reproduces gamma band activity (and also the synchrony of oscillations at separate sites that has been measured in the visual system) does include a component for subcortical inputs as well as both local and long-range cortical interactions (Wright, 1997 [315]; Wright & Liley, 1995 [314]).
In any case, whatever the origin of the 40 Hz oscillations that can be recorded at the scalp, it seems that whenever a human subject reports conscious experience, be this in the context of a dream or of wakefulness, 40 Hz oscillations are present. Conversely it is likely (though not at this stage proven) that whenever 40 Hz oscillations are present there is some degree of conscious experience. This suggests that at the very least, 40 Hz neural oscillations are somehow associated with conscious experience.
The major biologically important difference between the states of waking and sleep is that in sleep, external sensory stimuli are prevented from affecting the cerebral cortex and thus from entering consciousness. The mechanisms by which this block is imposed are known in some detail.
The brain region involved is the thalamus. The thalamus is perfectly placed for the role of regulating access of sensory stimuli to the cortex, because it is the one structure through which the neural pathways of all sensory systems (except the olfactory system) must pass on their way to the cortex. Not only are sensory messages not transmitted through the thalamus in sleep, but the state of the thalamus that blocks sensory traffic also causes major slow rhythmic activity in the cortex, of a sort which would overwhelm any of the smaller, faster and more complex sensory EEG patterns which did escape thalamic blockade.
The cellular mechanisms by which the thalamus either transmits sensory information or puts the cortex into a sleeping state are remarkably well worked out, but quite complex. Readers not particularly interested in neurophysiology could profitably skip the following (italicised) paragraphs without losing too much of the plot.
Thalamic neurons have two distinct modes of firing, depending on their resting membrane potential (Steriade & Llinas, 1988 [273]). During the waking state, they are relatively depolarized. In this condition they can be influenced by synaptic input from the peripheral sense organs to fire long trains of action potentials, which more or less faithfully transmit sensory information to the cortex (with some modification by attentional systems). During EEG-synchronised sleep, thalamic cells are basically more hyperpolarized. In this state they fire continuously in rhythmic bursts, for reasons which are described in the following sections. The blockade of incoming stimulus traffic from the periphery during this bursting mode is due to the fact that the neurons alternate between (a) being so hyperpolarized that incoming synaptic traffic does not bring the membrane potential close enough to threshold to cause action potentials, and (b) firing rebound bursts of action potentials after the cyclic periods of hyperpolarization (which also has the effect of blocking information that might otherwise be conveyed by patterned firing due to input from the sense organs).
The cellular mechanisms in the thalamus that underlie the states of light sleep and deep sleep are as follows.
The onset of sleep is signalled by sleep spindles. These are 1-2 second sequences of 7-14 Hz waves that wax and wane in amplitude so that each sequence has the shape of a spindle. A spindle occurs every 5 or 10 seconds during this stage of sleep.
Sleep spindles are generated by the reticular nucleus of the thalamus (Steriade, 1994 [277]). The reticular nucleus is a thin sheet of GABAergic inhibitory cells covering three sides of the thalamus. There are reciprocal synaptic connections between the reticular nucleus and each of the various other nuclei in the body of the dorsal thalamus. The nuclei within the body of the thalamus also send thalamocortical axons to various specific areas of cortex and receive corticothalamic connections back from the same areas of cortex to which they send axons.
The evidence that the reticular nucleus is the pacemaker of thalamic spindle oscillations (Steriade, 1999 [278]) is convincing. During spindling, reticular neurons show a slowly growing and decaying depolarization with superimposed spike barrages at a frequency of 7-14 Hz. This bursting pattern of firing is generated intrinsically, by interaction of the properties of six different types of ion channels in the cell membranes of the reticular cells. Reticular neurons are GABAergic inhibitory neurons, so the 7-14 Hz spike barrages they produce cause bursts of IPSPs in the cells to which they project in the main body of the thalamus. The successive IPSPs in each burst summate until they make the thalamic cell internally negative enough to activate an unusual type of calcium channel, which has the property that it is inactivated at rest and deinactivated by hyperpolarization. When this calcium channel has been deinactivated, the return to a more positive membrane potential after the spindle burst of IPSPs (which return is, incidentally, speeded up by the presence of an equally unusual non-inactivating sodium channel) can now open the calcium channels to generate a calcium spike. The calcium spike in turn triggers a so-called rebound burst of high frequency (200-400Hz) normal sodium action potentials in the thalamocortical cells. This burst of spikes is transferred to the cortex and it, together with its synaptic effects, show up on EEG records as sleep spindles. Complex, but effective!
The visibility of spindles in the EEG is assisted by the fact that sleep spindles in quite large areas of the cortex tend to be synchronised. This is because the bursts of IPSPs in the thalamic cells are synchronised, which in turn is because the pacemaking bursts in the reticular neurons are synchronised. Synchronisation of bursts in the reticular neurons may be because these are all interconnected within the reticular nucleus, by both dendro-dendritic synapses and gap junctions. Synchronisation may also be assisted by electric field-based volume conduction effects.
Spindling can be transmuted into the slower, continuous delta waves of deep sleep by a progressive hyperpolarization of the thalamic neurons. This is generated as sleep deepens because of a progressive decrease in firing rates of the cortical neurons that project excitatory synapses back to the thalamus. At a thalamic membrane potential more negative than around -70 mV, a non-specific cation channel called Ih is activated and the interaction between this current and the transient calcium current (It) referred to above in relation to the spindling sequence gives rise to the larger, slower oscillation of membrane potential that characterises delta sleep. In very deep sleep, the 1-4 Hz rhythm of delta sleep is further sculpted by an even slower rhythm with a period of between 2 and 10 seconds. This slow rhythm is disrupted by ketamine and thus probably involves NMDA receptor-mediated excitatory synaptic transmission, in some way that is not yet understood.
The above factors explain at a cellular level the generation of the various EEG rhythms of EEG-synchronised sleep. However they do not explain what leads the brain into a sleeping state in the first place, or what pitches it from the state of EEG synchronised sleep into the states of waking or REM sleep.
The main neuronal site involved in waking the brain up seems to be a small territory in the midbrain, at the mesopontine junction. Two groups of cholinergic cells in this region, plus the norepinephrine-containing cells of the locus coeruleus and the serotonin (5-HT)-containing cells of the dorsal raphe nucleus (all in the same restricted area) act conjointly on thalamic and cortical cells during wakefulness. In REM sleep, only the cholinergic neurons are involved and the locus coeruleus and raphe nucleus are virtually silent. This group of brain regions is sometimes called the ascending reticular activating system. The cellular mechanisms by which these activating neurons act are:
What activates the midbrain nuclei in question is controversial. One biological clock for the sleep-wake cycle is located in the suprachiasmatic nucleus of the hypothalamus, which has a direct link to the eyes (Kandel, Schwartz, & Jessell, 1991 [139]). As in the thalamic reticular nucleus, the cells of the suprachiasmatic nucleus are linked together by dendrodendritic synapses, which predisposes them to synchronous activity. This nucleus is spontaneously active and inactive in a circadian rhythm, which is entrained to the light-dark cycle in the environment by the retino-hypothalamic pathway. Whether this affects the thalamic reticular nucleus directly or through the ascending reticular activating system in the brainstem is unclear.
The three accepted states of consciousness are, as discussed above, waking, dreaming and sleeping. However, there is some basis for proposing the existence of a fourth state of consciousness, distinct from the other three. Whether this condition is in fact conceived as a distinct state unto itself, or as the basal state of waking consciousness, or simply as a transition state between waking and sleep, is a moot and probably not terribly important point. Whichever name you choose, this state is interesting from the point of view of understanding the varieties of consciousness.
By and large, the Western intellectual tradition emphasizes intentionality as a defining feature of consciousness (Rao, 1998 [232]). This means that waking consciousness is widely held to be necessarily of or about something. Such a view has been emphasized by Bretano, Husserl, Freud and Satre, it is the basis of phenomenology and it is espoused to some degree by most contemporary Western philosophers [e.g. Searle, 1993 [255]]. However the Eastern perspective on consciousness is quite different. Indian philosophy sees consciousness as basically non-intentional. Pure consciousness, or waking consciousness which is not of anything, is widely regarded as being not only theoretically possible, but attainable by ordinary persons. Indeed, various practical systems of meditation which have the immediate object of allowing this state to be reached by individuals on a regular basis have been taught for many centuries by Hindu and Buddhist masters in India, China and Japan.
EEG measurements tend to support the Eastern rather than the Western view on this issue. There are a number of distinct EEG conditions associated with what is subjectively reported to be a non-intentional state of consciousness, accessible by meditation or various other simple strategies. First the relevant EEG oscillations and their likely origins will be described and then the association between these EEG features and the subjectively reported state.
When a healthy adult human in a relaxed frame of mind rests with closed eyes, rhythmic activity with a frequency between 8 and 13 Hz can often be recorded from the posterior regions of the scalp, over the visual cortex. This so-called alpha rhythm can be plainly seen in raw EEG records and was first described 70 years ago by Hans Berger (Berger, 1929 [27]). Because it is normally blocked by opening the eyes and strongly suppressed during visual stimuli, visual memory tasks and visual imagery (Kaufman, Schwarz, Salustri, & Williamson, 1990 [141]; Michel, Kaufman, & Williamson, 1994 [191]; Salenius, Kajola, Thompson, Kosslyn, & Hari, 1995 [241]), alpha rhythm is generally considered to be the idling rhythm of the visual system. Consideration of whether or not this is true should probably be deferred until after the section on generation of the alpha rhythm, below.
Despite a large amount of work on the subject in the 70 years since human alpha was first described, the mechanism by which this major brain rhythm is generated is still controversial. Several hypotheses will be considered here. We might as well start with the most controversial one.
1. This is that the alpha rhythm is not generated by the brain at all, but by the corneoretinal potential, as modulated by 10 Hz physiological tremor in the eye muscles (Ennever, Lippold, & Novotny, 1971 [79]; Lippold, 1970a [166]; Lippold, 1970b [167]; Lippold, 1970c [168]; Lippold, 1971 [169]; Lippold, 1973 [170]; Lippold & Novotny, 1970 [171]).
The idea that the alpha rhythm is essentially an artefact has always been in the back of physiologists' minds and in fact worried Berger, the discoverer of human EEG, so much that he delayed publication of his findings for some time while he tried to convince himself that this was not the case. Today this hypothesis remains so threatening to EEG afficionados that most writers on electroencephalography either ignore it completely, or mention it briefly only to dismiss it without further consideration. In fact the idea, at least in the form outlined above, is supported by several experimental observations which can not be easily explained on the usual assumption that alpha rhythm is produced by the cerebral cortex under the recording electrode, and in my opinion it must be taken very seriously indeed. Observations supporting the hypothesis are as follows.
First, alpha rhythm is still present over an empty cranium after either total cerebral hemispherectomy (Cobb & Sears, 1960 [55]; Marshall & Walker, 1950 [180]; Obrador & Larramendi, 1950 [213]), or occipital lobectomy (Masland, Austin, & Grant, 1949 [181]). While it is conceivable that this simply results from far-field recording of sources in the intact cerebrum on the other side of the brain, it is difficult to reconcile this explanation with the observation that in some cases alpha rhythm over the section of empty skull was actually larger than that over the side with intact brain under it.
Secondly, it can be shown that in normal subjects tremor of the extraocular muscles is indeed related to alpha activity, as follows:
Thirdly, the corneoretinal potential (CRP) does seem to vary in concert with alpha amplitude. The corneoretinal potential is quite a large potential (about 80 mV) which can be recorded between the cornea and the back of the eye. It is usually measured by recording its movements, using the electro-oculogram. The CRP increases in the light and decreases in the dark, so it is considered that its cause is most likely the metabolism of visual purple. In light conditions, when the CRP is relatively large, alpha waves are also large. When the CRP is reduced by darkness, so is alpha amplitude.
Despite the rather compelling nature of the above arguments, however, it should be pointed out that even if the main generator of the alpha rhythm were to be accepted as being the corneoretinal potential, such large electromagnetic oscillations across the cortex would be very likely to induce currents and action potential firing in cortical cells anyway (see Chapter 7). Therefore the immediate cause of at least some of the alpha frequency oscillation normally measured over the occipital scalp is still likely to be brain activity at around 10 Hz.
2. The second hypothesis is that alpha rhythm in the cortex is generated by pacemakers in the thalamus. This idea too is somewhat controversial. Because sleep spindles have a frequency in the alpha range, experimental observations on the thalamic site of generation of spindles in barbiturate-anesthetised cats (Andersen & Andersson, 1968 [14]) were initially treated as revealing the mechanism of alpha rhythm generation. However, one major contributor to the field (Steriade & Llinas, 1988 [273]) argues that there are a number of differences between alpha and spindles: (a) their topography is different, with alpha-type rhythms being recorded from posterior areas around the occipital cortex and posterior temporal areas and spindles mostly from anterior areas (b) their structure is different, with alpha waves occurring in very long trains, while spindles are grouped in short sequences that recur periodically and in which the amplitude of the waves increases and then decreases in the shape of a spindle and the frequency decreases as amplitude increases, and (c) their behavioral concomitants are different, with spindling occurring during unconsciousness and being associated with depressed synaptic transmission through the thalamus, while alpha waves occur in the waking state and may increase in prevalence and amplitude during attentional demands. Thus alpha and spindling are now often considered to be quite different oscillations. My personal observations are that alpha spindles can certainly be observed in the EEG of humans who have their eyes closed but are obviously awake, so the situation seems to me to be less than clearcut at the moment.
In any case, the observable cellular mechanisms that presumably underlie alpha-frequency spindling in some way are outlined above, in the section on light sleep. It is probably fair to say that most neurophysiologists would take these findings as being all that is needed to explain the genesis of alpha-frequency spindles in the EEG.
3. Another class of ideas on this subject is represented by various mathematical models of the generation of the alpha and other EEG rhythms. These start by making fundamental assumptions about what neuronal groupings and interactions are important in rhythm generation. They then set up a simplified mathematical model of how these elements behave, assigning numerical values to the relevant parameters. In some cases the numerical values are drawn from anatomical and physiological measurements, but in other cases it must be admitted that they are essentially fudge factors, specified simply to make the model's predictions come out as desired. Finally, the models mathematically manipulate these parameters according to specified rules, in order to generate an oscillation in the system which is more or less close in frequency and behavior to the oscillation measured empirically in the brain. If they are useful, such models make specific, experimentally testable predictions about the characteristics of the rhythm whose generation they purport to explain.
There exist two well-developed mathematical models which deal specifically with the genesis of the alpha rhythm. The earlier of the two is that of van Rotterdam and Lopes da Silva (Rotterdam, Silva, Ende, Viergever, & Hermans, 1982 [239]; Silva, Hoeks, Smits, & Zetterberg, 1974 [262]; Silva, Vos, Mooibroek, & Rotterdam, 1980 [263]; Steriade, Gloor, Llinas, Lopes de Silva, & Mesulam, 1990 [274]). This model proceeds from the assumption that the alpha rhythm is generated partly by thalamic driving of cortical cells (as outlined in the section on the cellular mechanisms of spindling) and partly by intracortical processes. Because measured intracortical correlations or coherences in the alpha range are reported by this group to be larger than thalamocortical coherences, the model is biased towards intracortical generation of alpha. The model can be regarded as primarily a local model, in that it assumes only an infinite one-dimensional chain of pyramidal cells and interneurons, interconnected by collaterals and inhibitory fibers. When the neuronal chain is driven by a spatially and temporally noisy signal (from the thalamus), dispersive waves propagate along it at the frequency of the alpha rhythm. Anatomically derived size parameters are used and the wave consequently moves at approximately the speed that can be measured for the alpha rhythm. In other words, the model achieves a good fit with the temporal and spatial properties of the alpha rhythm at a millimetric scale. However this is at the price of linearising the system, imposing a grossly simplified anatomical structure and introducing a strength of coupling which is arbitrarily defined. Boundary conditions are open and waves outside the alpha range are not considered explicitly (Wright & Kydd, 1992 [313]).
The second mathematical model dealing specifically with the origin of the alpha rhythm is that of Nunez (Nunez, 1995 [211]). This model involves a local component but also brings global considerations into the picture, in the form of an influence from cortico-cortical association fibers. The model consists of two linear integral equations for global dynamics (representing the times taken for neural impulses to propagate in cortico-cortical association fibers) coupled to a third, generally nonlinear equation for local dynamics (which depends on the rise and decay times of cortical postsynaptic potentials). The model also introduces boundary conditions imposed by the skull, which leads to the prediction that alpha is actually a global standing wave. Usefully, this model makes two critical and falsifiable predictions - one of which has been criticized and the other falsified. It predicts (a) a negative correlation between brain size and fundamental mode alpha frequency and (b) a positive correlation between cortico-cortical long range fiber conduction velocity and fundamental mode alpha frequency. Evidence cited by the model's author in favor of the first of these predictions is attacked and the second prediction is experimentally falsified by Sergejew (Sergejew, 1997 [256]). Probably the main theoretical weakness of the model is that it fails to give subcortical (i.e. thalamic) influences any consideration at all, thus ignoring a large body of empirical data which says that they probably are important. Again, as with the Amsterdam model, the genesis of EEG oscillations with dominant frequencies other than that of the alpha rhythm is not dealt with.
A third mathematical model which generates alpha-frequency oscillations, along with all the other major cerebral rhythms, is that of Wright and Liley (Wright & Liley, 1995 [314]). This model has been discussed already in the section on the generation of gamma rhythms. It provides a useful bridge between the microscopic and the macroscopic scales in the understanding of EEG rhythms.
4. Current dogma among magnetoencephalographers is that alpha is generated in the visual cortex, mainly in the region of the calcarine sulcus (Chapman, Ilmoniemi, Barbanera, & Romani, 1984 [53]) and around the parieto-occipital sulcus (Salmelin & Hari, 1994a [244]; Vvedensky, Ilmoniemi, & Kajola, 1986 [302]). These conclusions are based on calculations of the site of hypothetical dipole sources which best explain the observed scalp recordings (proceeding from the assumption that the generator of alpha is somewhere in the cerebral cortex).
In summary, it can be postulated that one's views of the origin of the alpha rhythm are likely to be colored by one's professional background. More usefully, it can also be stated with some confidence that the labeling of all EEG oscillations that have a dominant mode in the frequency range 8-13 Hz as "alpha" is likely to be a gross oversimplification. There are probably a number of different "alpha" rhythms.
One form of spontaneous EEG activity in the frequency range 8-13 Hz is called the mu rhythm (Gastaut, 1952 [109]; Kuhlman, 1978 [153]). This can be recorded over the motor cortex near the top of the head. Its earlier name was the Rolandic rhythm. It is considered by magnetoencephalographers to be generated near the primary somatosensory hand projection cortex (Tiihonen, M.Kajola, & Hari, 1989 [289]) and is blocked by body movements and by tactile stimuli. The rhythm consists of two main frequency components, one around 10 Hz and the other around 20 Hz (but usually not exactly harmonic frequencies). The source locations of these as determined by MEG are usually about 5 mm more anterior for the 20 Hz than the 10 Hz rhythms (Salmelin & Hari, 1994b [245]). Human rolandic MEG activity has a close temporal relationship to peripheral muscular activity (Conway, et al., 1995 [56]; Salenius, Portin, Kajola, Salmelin, & Hari, 1997 [243]; Salenius, Salmelin, Neuper, Pfurtscheller, & Hari, 1996 [242]; Volkmann, et al., 1996 [301]) and there is some suggestion that the 20 Hz rhythm is predominantly related to the generation of motor activity while the 10 Hz rhythm is related to somatosensation.
Yet another 8-10 Hz rhythm is observable over the auditory cortex (Tiihonen, et al., 1991 [290]). This one is called the tau rhythm, because it is seen over the temporal lobe. It is dampened by sound stimuli, but not by opening the eyes or by movement. The rhythm has a similar spatial distribution to the auditory evoked magnetic field, with MEG-derived sources in the supratemporal auditory cortex.
The name theta is generally given to EEG oscillations in the frequency range between 4 and 8 Hz. However it is possible that as with the "alpha" rhythm, there is more than one functional rhythm underlying this general class. In fact there are at least three behavioral concomitants of oscillations in the theta frequency range that appear, at first sight anyway, to be quite different, and there are also at least two different sites of origin of such rhythms in the brain.
First, an EEG oscillation around 5-7 Hz occurs in some but not all subjects during the state of drowsiness between the restful alertness of the alpha range and sleep. Secondly, during mental calculation and intensive thinking, one 5-7 Hz MEG signal has been recorded from the frontal cortex (Sasaki, Tsujimoto, Nambu, Matsuzaki, & Kyuhou, 1994 [247]) and another from the hippocampus (Tesche, et al., 1995 [286]). The genesis of hippocampal theta has been intensively studied in rodents because of its importance in memory formation, but its cellular mechanisms are not yet clear (Stewart & Fox, 1990 [281]). Rhythmical EEG activity in the 6-7 Hz range over the frontal midline region has also been correlated with mental activity such as problem solving (Brazier & Casby, 1952 [33]; Ishihara & Yoshii, 1972 [128]; Mizuki, 1982 [197]; Mizuki, 1987 [199]; Mizuki, Takii, Nishijima, & Inanaga, 1983 [198]; Mizuki, Tanaka, Isozaki, Nishijima, & Inanaga, 1980 [196]). Some studies have reported being unable to induce the latter EEG rhythm (Niedermeyer, Krauss, & Peyser, 1989 [209]), but this finding was clarified by the observation (Takahashi, Shinomiya, Mori, & Tachibana, 1997 [284]) that individuals with frontal midline 6-7 Hz activity during drowsiness also had the same type of activity during mental tasks. It is a reasonable hypothesis that the mental tasks used in the studies showing theta rhythm during intensive thinking may have become either so boring or so fatiguing that they induced a state of drowsiness, so that the two frontal theta rhythms (which apparently have entirely different behavioral correlates) are actually one and the same.
A third, distinctly different kind of mental state underlies the so-called hedonic theta rhythm, which has been observed in young children during very pleasurable activities (Kugler & Laub, 1971 [152]; Maulsby, 1971 [184]). The response consisted of very strong 4 Hz activity in posterior and central areas, which was different from the 5-6 Hz rhythm produced in drowsiness by the same children. Such rhythms have not been seen in adults during sexual activity (Graber, Rohrbaugh, Newlin, Varner, & Ellingson, 1985 [116]), but little else has been done on the EEG correlates of pure pleasure. This seems strange, since one would think that pleasure should be a passingly enjoyable topic to study. Perhaps the Puritan work ethic so necessary for surviving the rigors of a scientific apprenticeship overwhelms such considerations. Or perhaps the problem is simply that granting agencies fail to see the benefits of a scientific understanding of happiness (which, if so, would seem to be remarkably short-sighted in view of the fact that clinical depression is one of the major psychiatric problems of our age).
The EEG of drowsiness will be discussed at this point because some authors consider that the meditational state is simply a finely-held state of drowsiness. The evidence given here suggests that this is not the case.
Drowsiness is hard to define exactly, but in general terms it is the state that normally occurs between waking and sleeping. Frequently drowsiness is defined in terms of a decrease, intermittency and finally cessation of the alpha rhythm. However about 11% of normal adults do not show alpha rhythms at all and most of those who do neither look nor report feeling drowsy up to 20 seconds after alpha dropout (Santamaria & Chiappa, 1987 [246]). Furthermore, there is considerable EEG variability in individual transitions from waking to sleep, both between subjects and between consecutive episodes of drowsiness in the same subject. All that being said, it is possible to make some generalizations about the EEG of drowsiness (although no mechanisms can be deduced from these, because of the frequent exceptions to them).
First, awake alpha is usually seen over the occipital lobes. During drowsiness, this tends to disappear and be replaced by temporal (10% of subjects) and/or centro-frontal (75% of subjects) "alpha" (Santamaria & Chiappa, 1987 [246]). As mentioned earlier, what is described as temporal alpha is probably actually the tau rhythm. If frontal alpha is in fact generated in the frontal lobes it may be associated with disappearance of thoughts, as occipital alpha is associated with disappearance of visual sensations and tau with disappearance of auditory sensations. Centrofrontal alpha appears in some drowsy subjects both when using a non-cephalic reference (balanced neck/chest) and with longitudinal bipolar montages, so it is not an artefact of false extension by the ear reference of posterior alpha to the anterior leads. Also, the centrofrontal alpha usually has a slightly slower frequency than the posterior alpha and there is usually a decrease in amplitude of the posterior alpha as frontal alpha appears.
Secondly, the amplitude of the posterior alpha can either decrease smoothly with only minimal slowing during the onset of drowsiness (about 30% of subjects) or it can actually increase, without spreading to the vertex (25% of subjects). One review considers an increase in alpha amplitude and a mild slowing to be the first EEG signs of drowsiness (Erwin, Somerville, & Radtke, 1984 [81]).
Thirdly, it is a general statement that slow EEG rhythms increase as drowsiness progresses, but the patterns with which this happens vary. One pattern shows posterior alpha decreasing in amplitude and regularity and becoming intermixed with irregular, moderate amplitude theta/delta oscillations. Sometimes in this pattern the theta and delta activity is present centrofrontally as well. A second pattern, which is very common, consists of runs of centrofrontal, moderate amplitude 3-5 Hz activity concurrent with awake posterior alpha. Various other patterns of slowing or disappearance of alpha and appearance of theta rhythms have also been reported. As well as biological variability, the particular electrode montage used is important in determining which rhythms are seen and which are not. Whatever the cause, the general picture is of a very non-homogeneous transition from waking to sleep EEGs.
The EEG of meditation, on the other hand, presents a much more homogeneous picture than that of "natural" drowsiness.
The two main meditation traditions which have participated in EEG studies to date are Zen Buddhism and Transcendental MeditationTM (a.k.a TM), a Vedic Hindu-based meditation technique popularized and essentially franchised in the West by Maharishi Mahesh Yogi. There are, of course, many other meditation techniques, but EEG recordings have been done mainly on Zen and TM practitioners. The meditation techniques used by these two traditions have a similar goal. This is a state which is variously described as very restful, silent, often blissful, in which the subject is awake (in fact often feels a heightened alertness), but has no thoughts or sensations, except sometimes a pervading feeling of objectless bliss or ecstasy. These subjective descriptions already differentiate the state from drowsiness.
The techniques used in Zen and TM are in some ways similar, but in other ways distinctly different. Zen meditation involves concentration on some object such as a single thought or sensation, the subject's breathing, or various purposely paradoxical, counterconceptual problems called koans. Beginners usually find difficulty in maintaining concentration. Then, as the subject's ability to concentrate improves, various odd bodily sensations may occur. With perseverance, extraneous thoughts and sensations become less intrusive and fade away completely. Eventually even the object of concentration fades away and a wakeful state variously called "nirvana" or "satori" is achieved. Zazen, the most usual, sitting form of Zen meditation, emphasizes absolute physical stillness in a particular posture during meditations. Sleep and intruding thoughts are positively rejected. The practice is done with the eyes open. There are various sects of Zen Buddhism, and even within one sect the details of meditative procedures depend on the particular Zen master and his estimate of the individual student. A flash of satori can occasionally be achieved instantaneously, given the right mental and physical preparation and a talented teacher, but it is accepted that generally many years of dedicated practice of the technique may be necessary before the desired state can be reached reliably.
In contrast, the technique of TM produces reports of some subjects' achieving the desired state of nothingness after only a few days of practice. TM involves silent mental repetition of a word called a mantra. The main feature of the technique is supposed to be its effortlessness. Neither sleep nor thoughts are rigidly rejected, physical stillness is not particularly emphasized (although it does occur spontaneously, right down to a dramatic slowing of breathing) and the technique is done in any comfortable upright sitting position, with the eyes closed. The immediate aim is a state where spontaneous thoughts cease and finally even the mantra disappears and one is left with a non-intentional state of consciousness called "samadhi" or "pure consciousness". This has been characterised on the basis of physiological measurements as a wakeful, hypometabolic integrated response (Jevning, Wallace, & Beidebach, 1992 [132]). Thereafter, repeated experience of this state is said to produce a variety of permanent "higher" states of consciousness. Unlike Zen, the TM system has notably emphasized standardization of teaching methods, which makes it a particularly attractive candidate for scientific study.
The EEG concomitants of meditation in both traditions are similar. The largest study of the EEG of Zen meditation so far has been done by Tomio Hirai (Hirai, 1974 [127]), who studied 48 monks of the Soto and Rinzai sects. The meditation experience of these subjects ranged from 22 to 55 years. A good correlation was seen between an independent assessment by the Zen masters of individuals' advancement in the Zen sense and the EEG changes occurring during meditation. The basic EEG correlate of meditations that were deemed to be successful was the appearance of alpha rhythm in the eyes-open state (although it should be noted that the monks in the photographs shown by Hirai appear to have their eyes about half closed, as a result of looking downwards at a point about a meter in front of them). The alpha oscillations were more prominent in frontal, central and parietal recording sites than occipital. In those monks with relatively more years of meditation experience the alpha oscillations became larger and slower, particularly at more frontal sites, as a particular meditation progressed in time. In a few of the most advanced monks, high voltage rhythmical theta waves then appeared at all recording sites. These theta waves had characteristics different from the theta waves seen in drowsiness. They tended to be of higher amplitude than the low-voltage drowsiness theta. Furthermore, in the drowsy state an audible click will stop theta activity and elicit alpha instead. However a click presented during meditative theta elicited a blocking response but no alpha - the EEG was simply desynchronised briefly by the click and the theta rhythm reappeared after 2-3 seconds. Drowsy-type theta was sometimes seen in younger monks during meditation, but this was clearly distinguishable from the theta seen in deep meditation by the advanced monks, both electrophysiologically and by introspective report. Breathing slows down considerably during the meditations of experienced Zen practitioners, but this is a voluntary rather than an involuntary effect, in that the training emphasizes breathing softly and concentrating on breaths.
TM produces a remarkably similar EEG profile to Zen meditation. Overall, about 40% of meditation time is spent in sleep as defined by the standard EEG sleep criteria. This almost certainly depends simply on how sleep-deprived the particular individual is at the time of meditation. In meditations where sleep does not intervene, the general pattern with TM is that alpha rhythm increases in amplitude, slows down in frequency and extends to anterior channels at the start of the meditation (Banquet, 1973 [21]; Wallace, 1970 [303]). No obvious correlations are reported between this period and any particular subjective experience. In a second stage, bursts of theta frequencies, different from those of sleep, diffuse from frontal to posterior channels. Periods of theta correspond closely with periods of involuntary breath suspension for up to a minute (NB:TM does not involve any active breath control, or indeed any attention to the breathing at all). During these periods the subjective experience is uniformly reported as peaceful, comfortable or pleasant, with no thoughts but a full, "expanded" awareness of perfect stillness (Farrow & Hebert, 1982 [85]; Hebert & Lehman, 1977 [123]). This experience is identified by the TM culture as pure consciousness, or transcendence. The end of such periods is signalled by a burst of beta frequency EEG activity and resumption of normal breathing. Particularly during the first part of the pure consciousness periods, coherence between the EEG recorded at different electrode sites, particularly in the theta band, is reported to be very high. However a sharp rise in basal skin resistance is also reported to occur at these times and this together with reference electrode effects might well result in some artefact with regard to coherence measurements.
Thus overall, with two different techniques of meditation, we find that the desired effect of a state of pleasant or even blissful thought-free, non-intentional consciousness, is correlated with the presence of a distinct type of theta frequency EEG activity. It is tempting to identify this with the hedonic theta seen in children, but more work is necessary in this area before any conclusions can be drawn. The major difference between TM and Zen techniques is that this state seems to be considerably easier to achieve with TM than with Zen meditation.
Since the state is relatively easy to achieve during TM meditations, it is regarded in the TM culture as merely the baseline for the development of three permanent "higher" states of consciousness. This is in sharp contrast to the situation with most other schools of meditation, where the experience of samadhi or satori (a.k.a. pure consciousness) is regarded as the ultimate goal, to be reached by only a few of those who devote their lives to meditative practices. The Vedic tradition underlying TM says that with repeated meditation practice, the first of three higher states can be stabilized, in which pure consciousness is spontaneously experienced continuously, not only during meditations but during waking, dreaming and even deep sleep. The term "witnessing" is used to describe this state because transcendental consciousness is experienced to be a non-changing level of awareness that serves as a peaceful inner observer or "silent witness" to the active changing states of waking, dreaming and sleep. Witnessing during sleep apparently has some similarities to lucid dreaming, which is a condition in which one is aware that one is dreaming and can control the dream (LaBerge, 1985 [157]). However in the case of witnessing during dreaming, the inner observer takes no part in the dream but simply calmly looks on, as it does also during deep sleep. Remarkably, TM practitioners who report clear experiences of witnessing in sleep do show 6-10 Hz theta-alpha activity simultaneously with delta activity and decreased chin muscle activity during deep sleep (Mason, et al., 1997 [182]). Control subjects did not show such EEG activity.
The relationship of meditative theta activity to the theta activity seen in concentration on mental tasks is unclear, but on the face of it, these studies on meditation do seem to provide both objective and subjective evidence for the existence of a fourth state of consciousness, which correlates with the presence of EEG oscillations in the theta-alpha range.
The experimental work cited in this chapter shows that the existence of the three usually recognized states of consciousness - waking, dreaming and deep sleep - can be accounted for or explained in terms of the hypothesis put forward here. That is, there is good correlation between the presence of certain global electromagnetic oscillations in the brain and the presence of associated states of consciousness.
Specifically, the available evidence supports the postulate that gamma frequency oscillations (20-80 Hz) are the carrier wave of the various modalities of waking and dreaming consciousness (such as olfactory, auditory and visual conscious experience). The conscious contents of these modalities are very likely to be 3-dimensional spatial modulations of the carrier waves, as will be described in later chapters. Deep sleep is associated with electromagnetic oscillations in the frequency range 1-4 Hz and lighter sleep with slightly higher frequency oscillations. There is some evidence suggestive of the existence of a fourth state of consciousness, which may be described as pure, or non-intentional consciousness. This may possibly be identified with electromagnetic oscillations in the 4-10 Hz range.
The major area concerning states of consciousness on which the electromagnetic theory as it stands does not provide much explanatory power is the difference between waking and dreaming. These states can not be readily distinguished by gross examination of brain-generated electromagnetic patterns at this point. However, since both waking and dreaming do involve conscious experiences, this defect may be seen as one of detail rather than of major substance. From one point of view, the distinction between waking and dreaming has more to do with the specific contents of the conscious experiences than anything - and our theory has so far been at least partly successful in showing why the conscious contents of the dream state differ from those of the waking state (in terms of the prevention by the thalamus of sensory input from reaching the cortex).