One issue in particular has been weighing on a number of people with regard to our work. It is the question of why a single protocol should be so effective for such a variety of conditions, and why a particular virtue seems to attach to the use of bipolar training, a tactic that has been abandoned by many in the field who have made the transition to QEEG-based training. This issue has come up again recently, so this is not a bad time to discuss it. A secondary issue is why inter-hemispheric training should hold such special virtues for us, but that issue can await its own individual treatment.

First of all, it needs to be recalled that all of the early work of Barry Sterman and Joel Lubar was done with bipolar placement, which is characterized by the fact that both active leads are placed on the scalp over cortex, as distinguished from referential placement in which one active lead is placed on a “quasi-neutral” site such as the ear.

Secondly, it needs to be kept in mind that the claim of broad efficacy for a single reward frequency also has a long pedigree. After gaining a foothold in the management of seizures and ADHD, SMR training was found to offer clinical effectiveness for a host of conditions. These conditions included anxiety and depression, bruxism, motor and vocal tics, headaches and other pain syndromes, and a variety of other conditions. Throughout this evolution of clinical practice, there was essentially one constant: the SMR reward frequency of nominally 12-15 Hz. In our case we balanced the soothing and calming aspect of SMR-training with the more activating low-beta reward frequency of 15-18Hz. But nevertheless, essentially all of our work was done with these two reward bands for a number of years.

Meanwhile, the more prominent part of the neurofeedback field in the early days also focused on a single reward band, namely the alpha band. So essentially all of neurofeedback concerned itself with what could be called the cortical resting rhythms, the alpha rhythm in the case of the visual cortex, and the sensorimotor rhythm in the case of the somatosensory and somatomotor cortex. Sterman found that the SMR bursting activity actually extended from some nine to nineteen Hz in the cat, so the 15-18Hz band could legitimately also be considered part of the sensorimotor rhythm. He certainly considered it to be so when he first investigated training in that band.

At some point we prevailed upon our software writer Ed Dillingham to give us a fifteen Hz filter as a “vernier” on our two favorite frequency bands, in the event one was too activating or the other too calming. This helped to confirm that we were in fact working on a continuum in the arousal domain for which the reward frequency was the vernier. In time we came to occupy the whole “neighborhood” of the standard bands for particular purposes for different people. What remained constant throughout, however, was the fact that people trained at characteristic frequencies. These were just not uniform among clients. Every person had their “comfort zone” for training, and it was quite clear that people did not have different target zones for different problems. Whatever could be accomplished at one training site could be accomplished with essentially one reward frequency.

So the observation that one reward frequency serves the purpose of general improvement in self-regulation status seemed to be confirmed–only the frequency was not uniform among clients. We somewhat informally carried in our heads the idea of a rather fixed relationship between a person’s functioning in the arousal domain and the frequency at which the respective brain likes to train. We were experiencing the whole range of dysfunction from the fogginess of under-arousal to the agitation of high-arousal conditions within the range of SMR to low beta frequencies, i.e. in the range of 12-18 Hz.

It is not surprising, therefore, that we imagined a kind of “arousal scale” that would map into the frequency scale in this regime. It might end up being a rubber ruler, but perhaps a useful ruler nonetheless. We certainly did have the idea that most people were being accommodated within the new “standard range” of frequencies of 12-18Hz. Of course there were people that chose to walk away from our training, and there were some conditions that we did not regard as trainable with our methods of the day. It was the inter-hemispheric training that broke the mold in which our thinking had congealed. The training was obviously stronger in its effect for some conditions than the protocols we had used up to that time. They also had to be optimized more carefully. At the same time we were being challenged more to address the autism spectrum that manifestly did not train well with the standard bands.

We were driven to lower and lower reward frequencies both in pursuit of the autism spectrum and of Reactive Attachment Disorder. Cerebral palsy confirmed the trend. And our work with the most sensitive responders—fibromyalgia patients, migraineurs, etc. confirmed the expansion of our horizon into the forbidden frequency bands of “theta” and even “delta.” But the two earlier observations still held: there was one reward frequency per person, and that reward frequency had essentially nothing to do with the particular symptom expression. Also a certain ambiguous correlation remained between training frequency and presumptive arousal level, in that the most highly over-aroused clients trained at the lowest frequencies. But our rubber ruler that ostensibly mapped arousal level into reward frequency had taken on a Daliesque quality. Even its basic topology was in question.

When we then took our new-found observations from inter-hemispheric training back to lateralized training, we found a similar frequency dependence. We had simply missed it before, as we were trying to make everything fit into the 12-18Hz, or at most 8-18 Hz regime for which our rubber ruler had been trimmed. A consistent and more comprehensive story now emerged in which inter-hemispheric training took up its rightful place in the remediation of the most severe, the most global, and the most deep-seated dysregulations.

The rather comprehensive efficacy of the inter-hemispheric training is a remarkable phenomenon that still requires explanation–and still invites skepticism from those who have never tried it. The elegant simplicity of a single-channel bipolar approach should be matched by a similarly elegant, simple explanation. We already know that we cannot draw upon particular characteristics of representative disorders for an explanation because the training is not diagnostically specific. We also know that the training does not lead to systematic alteration in the stationary EEG properties at the training site and in the direction of the protocol.

Of course in order to make such a statement one has to be able to stipulate the change we expect from the protocol. In its early implementation by Lubar and Sterman, it was referred to as amplitude training, by extrapolation from the early work on cats. And that interpretation has gained a certain validity simply by having survived this long. But matters actually got murky in several respects as neurofeedback transitioned from cats to people. The cats actually exhibited synchronized SMR bursting activity that could be readily distinguished from background activity, and which was localized to cat sensorimotor cortex. Hence it was easy to implement a referential placement where only one electrode reflected the SMR bursting activity, and the other active electrode in the bipolar pair remained on neutral ground.

Matters are very different in work with human subjects. First of all, we don’t observe the obvious synchronized SMR bursting activity that rises significantly above background in the waking human EEG. Secondly, when a bipolar placement of C3-T3 is employed, as Sterman and Lubar both did in their early work, there is no obvious prediction as to what might happen. A general increase in SMR amplitudes with training on the sensorimotor strip will also yield an increase in average amplitude seen in C3-T3, but we cannot be more specific than that. Somewhat problematically, such increases in amplitude were not seen consistently, and were not consistently related to clinical change. It was this lack of obvious correlation that was perhaps most responsible for the neglect of Sterman’s findings originally. Just as we witness in the present day, critics have a propensity to magnify any conceivable flaw in a theory they find unpalatable, and here one was served up for them on a platter. It did not help that Sterman himself looked to the systematic increase in SMR amplitudes for confirmation of his model.

In consequence, we must be in a position to explain how substantive clinical change can be consistent with relative little change in the parameters being challenged in the training. The most obvious answer is that the EEG is actually a highly dynamic and state dependent measure. Hence, a substantial degree of change may quite simply escape our notice in the face of the intrinsic variability of the signal, on the one hand, and our inability to specify the state in which measurement occurs, on the other. This explanation may be quite sufficient, actually. But we would like to propose another: Neurofeedback may in fact be taking place in the context of a “tightly constrained” system. That is to say, the brain takes great pains to maintain the integrity of its neuronal assemblies, and any attempt to alter them will be massively resisted. To challenge the organization of the neuronal assemblies in the time and frequency domains is not so much to alter these assemblies as it is to strengthen the mechanisms by which they are organized.

An analogy might help here. In our elementary school playgrounds we can all recall contests in which children would ride on the shoulders of another and try to knock each other down. The attempt to knock the adversary off the shoulders of his host would be massively resisted by the host, but the amount of effort involved would not be obvious to an observer. As long as there was not complete breakdown, one would observe that the center of gravity of the person on top would be roughly above the center of gravity of the person below. Only with incipient breakdown would a deviation become evident. Thus it may be with the brain. On the one hand it may yield to our neurofeedback challenge, but the rest of the brain becomes aware of this intrusion into the integrity of its affairs, and it mounts a response. The result is that a system with decent functional integrity will not show substantial alteration in the presence of a neurofeedback challenge under any circumstances. It is the act of resistance that ultimately effects the strengthening of the regulatory apparatus. And if that is so, then we should train where we have the greatest leverage to engage the resources of resistance, not necessarily where the functional deficit is most manifest.

So the essence of neurofeedback in this model is not to direct the brain toward a particular endpoint of the training process but rather to act as a kind of “provocateur,” as a subliminal challenge to the regulatory regime. We call this the challenge model. And in that event, what is being challenged in the single-channel bipolar montage? In the moment of training, we engage both the relative amplitude and the relative phase of the EEG at the two active sites. Both variables are in play. Neither has priority in principle over the other. But both are engaged in such a way as to move the brain toward a differentiation of the EEG at the two sites. The direction of our challenge is toward increasing amplitude disparity and phase disparity. Up to now, the discussion has been entirely about amplitude, to the virtual exclusion of the issue of relative phase. Not only does the issue of phase need to be part of the discussion, but it may turn out to be dominant in much of our clinical work. This matter cannot be resolved in the abstract. We must look to how nature actually behaves in order to make more specific statements about how the neurofeedback engages with the organization of our neuronal assemblies.

In our next newsletter, we will set the stage with a more mathematical treatment of the issue of phase in neurofeedback.

This entry was posted on Wednesday, June 7th, 2006 at 8:06 pm and is filed under Efficacy, Paradigm, Professional Issues, Scientific. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.