Hemispheric Specificity of Training: Spatial Dependence of Protocols
The clinical data reviewed below are supportive of the view that the training exercises the two hemispheres specifically, and differentially. Cumulative clinical evidence in our offices has also reinforced the view that referential training near C3 and/or C4 is generally the most effective. Small displacements from these sites laterally from the midline along the coronal plane seem to have a minor effect on the training. Small displacements in the horizontal plane, on the other hand, change the quality of the training more significantly in our experience. Hence the training sites have been determined by a process of local optimization (i.e., small spatial displacments), rather than of global optimization. For some applications (principally to the instabilities), T3 and T4 have been found preferable to C3 and C4, respectively.

With a large amount of clinical data at our disposal (several thousand cases), a picture has emerged that the EEG training addresses the specific failure modes of each hemisphere. If a particular disorder could be associated more directly with one hemisphere than the other, it might give us a clue as to what part of the brain might require redress. Such a connection would then imply a unique, differential protocol. We found this to be true, and a number of disorders began to yield to assignment to one hemisphere or the other. Then, using a process of "local optimization" both in terms of spatial location and selection of the reward frequency band, a training strategy emerged which has gained considerable 'stability' from the effort at continual refinement, and what may have started out as mere clinical impressions have gradually been reinforced to the point at which they now constitute a defensible training strategy. The principal hallmarks of the strategy are as follows:

1. There appears to be a certain simplicity and directness attached to training along the sensorimotor strip.

2. Training away from the midline appears to yield stronger and more hemisphere-specific training effects, than training at Cz.

3. There is a distinct predominance of the need for up- regulation of the left hemisphere, using beta training (nominally 15-18 Hz), and a corresponding predominance of the need for down-regulation of the right hemisphere using SMR- training (nominally 12-15 Hz). Frequently, the need for both exist within the same individual. (This frequency dependence is addresses further below.)

The apparent advantage of training at the sensorimotor strip for most of the conditions discussed is consistent with the early Sterman hypothesis, since amply validated, that what is being trained is the degree of rhythmicity of the thalamocortical regulatory circuitry. And whereas the rhythmic EEG activity observable anywhere on cortex is traceable to these thalamically-mediated regulatory functions, the primary sensory areas of cortex are perhaps the most direct access we have to them. Specifically, the highest cortico-thalamic fibre-density is to be found in the primary sensory areas of cortex (and also in projections to the frontal lobe). Historically, most of the EEG biofeedback training that has been done has focused on the primary sensory regions.

Our continuing observation over a large clinical population of the need for up-regulation of the left hemisphere and down- regulation of the right can be explained in terms of the specific way in which the two hemispheres fail, or disregulate. The work of Malone, Kershner, and Swanson, et al, (1994), provides us with a detailed neurophysiological model which explains this hemispheric laterality in training effect. In this model, it is proposed that the left hemisphere (in collaboration with the frontal lobe) manages tonic activation for the conduct of intellectual and motor tasks, and for the maintenance of vigilance over time. This activity is preferentially under the management of the neuromodulators dopamine and to a certain extent acetylcholine. The right hemisphere, by contrast, manages phasic arousal for maintenance of sensory system readiness to receive and process new inputs from any source. This system is predominantly under the management of norepinephrine and to a certain extent serotonin.

The model, as applied to ADD, which will be discussed further in the coverage of our clinical outcomes, reveals ADD to be a problem of underactivation of the left hemisphere, principally involving dopamine, and of overarousal of the right hemisphere, principally involving norepinephrine. Hence, neither the sequential processing of intellectual or motor tasks, nor the deployment of resources responding to new incoming stimuli are well managed. The efficacy of Ritalin is attributed to a dual influence, the up-regulation of the dopamine system and the down-regulation of the norepinephrine system. In a kind of parallel or equivalent model, ADD of the inattentive subtype is addressed with higher frequency left hemisphere training (central and possibly frontal) and ADD of the impulsive subtype is addressed with lower frequency training of the right hemisphere (central and possibly the parietal region as well). A mutual consistency thus emerges between the claims of EEG biofeedback and psychopharmacology for ADD. The Tucker and Williamson (1984) model lays a credible foundation for the general claim that the two hemispheres need to be specifically and differentially addressed in the training, just as they are pharmacologically. Recent clinical work has led to further refinements of the principal protocols so that they now incorporate frontal and parietal training with bipolar placements that combine left central with prefrontal sites (e.g., C3-Fpz), and right central and parietal sites (e.g., C4-Pz).

These latter refinements specifically challenge communication loops between the selected sites. When a bipolar montage is used, then the reinforcement promotes an anti-phase relationship between the two sites. This may be counter- intuitive. It has been shown (Rappelsberger, 1994) that when distant cortical locations communicate with one another, they come into greater synchronization in the process. Why then would one wish to train these sites to reduce the prevailing degree of synchrony? The only justification that really counts is that this has been found effective empirically. The theoretical justification is to be found in the 'regulatory challenge' model of EEG biofeedback. The biofeedback reinforcement takes the brain momentarily out of its prevailing equilibrium, to which it then wishes to return. It may not matter in first order whether the disequilibration occurs in one direction or the other. Improved regulatory function may eventuate in either case.

It may now be possible to generalize the Malone model to other conditions. Just as there are left hemisphere and right hemisphere aspects of ADD, the same may hold for affective disorders of depression and anxiety (Goodwin,1990). The left hemisphere aspects of depression and anxiety may have to do with anticipatory activity, planning, ruminating, perseverating, worrying. The right hemisphere, by contrast, may harbor the non-rational, more catastrophic aspects of depression and anxiety, namely fear, panic, agitated depression, and suicidality (Heller, 1997). With a spatially localizable technique at our disposal, hemispheric specificities have been confirmed with EEG training not only for ADD, cognitive function, anxiety, and depression, but also for pain syndromes, sleep disorders, eating disorders, endocrine and immune system disorders. Laterality turns out to be one of the key organizing principles for the evolution of protocols.

The Protocols' Frequency Dependence

Protocols used for EEG biofeedback training of the 12-19 Hz band, are essentially derived from Sterman's seminal work with seizures. The 12-19 Hz region was originally identified as being prominent in the bursts of sensorimotor rhythm of the cat (Sterman, 1969). Subsequently, operant conditioning of the cat EEG was restricted to the peak frequency range of this distribution, 12-15 Hz (Sterman, 1970). As additional work was undertaken with human subjects, the 15-18Hz band was also investigated in one study (Sterman, 1978). In the following, we will refer to training with the lower frequency (12-15 Hz) and higher frequency (15-18 Hz) bands. The lower frequency training has also been colloquially referred to as "SMR" training, for historical reasons, and the higher frequency as "beta" training. These terms have become commonplace through clinical usage, even though we are dealing with only a subset of the entire beta band, which extends from 12 or 13 Hz to 35 Hz.

As we entered the field in 1985, we were aware only of the work of Barry Sterman, Joel Lubar et al., Michael Tansey, and Margaret Ayers with respect to the beta/SMR training. Joel Lubar et al. utilized both bands in the treatment of ADD (Lubar, 1984). Michael Tansey restricted himself to rewarding the frequency region centered on 14 Hz (Tansey, 1990), and Margaret Ayers used almost exclusively beta training (Ayers, 1993). In terms of electrode placement, Lubar et al. were typically using left-side training with bipolar placement near the sensorimotor strip, not deviating far from what Sterman had originally employed (C3-T3). Tansey exclusively used an electrode placement on the supplementary motor area, with a large-area contact that covered the space between Cz forward toward Fz, and also extending partially toward Pz. Margaret Ayers used C3-T3 placement almost exclusively, except when either symptomatology or EEG phenomenology indicated a need for right-side training at C4-T4.

All of the above protocols were accompanied by inhibition of low frequency activity, typically 4-7 Hz (called "theta" in the following). In the case of Michael Tansey, the information regarding excessive theta amplitudes was verbally communicated to the client. Additionally, Sterman and Lubar provided for inhibition of high-frequency activity in the region above 20 Hz.

Out of the work of these four pioneers, our protocols evolved in several stages. First, placement was changed from bipolar to referential to the ipsilateral ear, in line with a general trend within the field toward referential montage. Secondly, Cz placement was evaluated for the low frequency training on the basis of Tansey's work. For more than a year, most of the training was conducted at either C3 with the higher frequency band ('beta'), or at Cz with the lower frequency band (SMR), using an A1 reference. Excursions to C4 were, if needed, based on our early understanding of issues of laterality or in cases of localization of deficits to the right side (as in seizure disorders, head injury, and stroke). Over time, as we became more experienced and our understanding of the hemispheric specificity of certain aspects of cortical disregulation became clearer, it was observed that the C4 training was typically most effective with the lower frequency training, and that often stronger, more specific results were obtained than at Cz. Eventually, the predominant protocols became C3-beta and C4-SMR. Some frontal and parietal training was used as well to address specific issues.

Though early protocol selection was based upon the prior research work, it soon became necessary to devise methods of assessment (to be discussed later in this piece) that would assist us in teasing out which of these protocols were most appropriate for the client. But if the judgment turned out to be mistaken, then there was always the option to make an early change in protocol. If the choice was appropriate, then a different protocol might be used later to address residual issues.

It was observed also that if one persisted with the use of a single protocol, then eventually certain adverse symptoms could develop which called for compensatory training. Thus, with left-side training, ultimately client reports might indicate the need for right-side training, and vice-versa. Subsequently, more refined clinical skills led to an earlier integration of the secondary protocol into the training for optimization. This compensatory training led to the appreciation that in addition to addressing the specific failure modes of each hemisphere we really had to also achieve, or maintain, hemispheric balance. Symptoms could often be attributed to the inappropriate inhibition of one hemisphere by the other, or inappropriate disinhibition. This was most directly demonstrated when a left-side seizure focus was also favorably influenced by training the contralateral placement. But the principle has proved to be valid broadly.

At the present stage of evolution of protocols, there has effectively been an integration of the C3-beta and C4-SMR protocols, which are both used with the majority of clients, generally within the same session, and the balance between them is titrated on the basis of symptom response. Assessment is then a matter of determining the client's physiological response characteristics, and the particular vulnerabilities expressed in their symptoms. In this appraisal, established clinical categories (from the DSM-IV) are only approximate guideposts. Whether diagnostic criteria are met in one respect or another is therefore irrelevant to the clinical burden. At least 80% of clients have been treated with this combination of protocols and this combination alone. The data reported in the following were obtained over the past eight years with the above protocols or derivations therefrom.