The Great Divergence

by Siegfried Othmer | August 6th, 2009

It goes without saying that mainstream thinking about neurofeedback to date has been mistaken. The original attempts at replication of Kamiya's work on alpha training for anxiety were misguided in their methodology and in their conclusions. The rejection of Sterman's and Lubar's collective body of work was a blunder of the first magnitude.It goes without saying that mainstream thinking about neurofeedback to date has been mistaken. The original attempts at replication of Kamiya’s work on alpha training for anxiety were misguided in their methodology and in their conclusions. The rejection of Sterman’s and Lubar’s collective body of work was a blunder of the first magnitude. The continued dismissal of the claims of neurofeedback in the face of mounting evidence is indefensible. On the other hand, nothing here really surprises. A paradigm shift of such magnitude will be resisted by the mainstream on all fours. Everything has gone true to the historical pattern with respect to scientific revolutions.

Far more calamitous for the subsequent history of the field has been that the pathology of mainstream thinking was carried right into our midst, and this has divided our common purpose. On one side have been those trying to uphold the standards of science as they had come to know them, and on the other were the scientist-practitioners who saw niches opening and went about to exploit them. They in turn attracted clinicians who were guided mainly by what did and did not work for their clients. It turned out that nearly everything worked to one degree or another. Of course the clinicians were bounded in their initiative by instrumentatal limitations, so in effect the field was guided by standard-model scientists on the one hand, and instrument developers and scientist-practitioners on the other. The field quickly Balkanized, but the chief division throughout has been between those with an essentially Platonic, model-driven and left-brained orientation to the work, and those who took a more tactical and exploratory approach. Sensitive clinicians, for their part, mostly came to the task with a functional right hemisphere and an engaged limbic system. The academic scientists were implementing a kind of feedback that mechanized the key principles as they knew them, whereas the new developers and their clinicians were being shaped by clinical experience into something quite different.

This has led to an ongoing experiment of nature, one in which the adherents of the “standard model” have been continually upstaged by the clinicians and their respective developers in terms of clinical results. This should never have been an affront to the researchers, because the clinicians are manning the frontier in great numbers, with lots of data-gathering opportunity. Their work should have been attended to as assiduously as ladybugs tend to their aphids. Instead the clinicians were always met with the critique that their observations did not rise to the level of science. That was true, of course, but the conversation needed to start somewhere. What if Gottlieb had dismissed Joel Weisman in the same fashion? The discovery of AIDS might have been delayed for years. Fortunately, he had his own case to back up what Weisman was telling him.

The standard model can be characterized by a number of essential pillars:

1. Neurofeedback is a technique with specific efficacy for specific conditions
2. Hence, the scientific case for neurofeedback needs to be established independently for each diagnosis
3. The classic diagnostic categories delimit those who may benefit from neurofeedback
4. Efficacy of neurofeedback must be established via group studies that include control groups and preferably use blinded designs
5. Within the diagnostic categories, QEEG criteria further delimit those who may benefit from neurofeedback
6. QEEG criteria should also be used to document the benefit derived from training
7. Targets for neurofeedback training should be selected on the basis of deviations seen in the quantitative EEG
8. The scientific paradigm underlying neurofeedback is that of operant conditioning
9. Hence, the training is fundamentally cognitively mediated
10. Therefore, rewards should be discrete, relatively rare, and they should allow for a refractory period for consolidation consistent with learning theory

Several subsidiary implications have been drawn from the above cardinal principles:

1. Young children are deemed untrainable for lack of cognitive maturity
2. The elderly are deemed incapable of responding to neurofeedback, presumably for lack of residual brain plasticity
3. In selecting ADHD children for training, other diagnoses such as anxiety and depression need to be screened out
4. An EEG criterion can and should be used to select for ADHD children who are responsive to neurofeedback

When one proceeds from a doctrinaire position, it is difficult to gather the evidence that would contradict the hypothesis. For example, if one starts out convinced that young brains cannot respond to neurofeedback, one would never seek the opportunity to test that proposition because it would be unethical to do so. If one believes that children need to be pre-screened to a certain criterion, how would one ever find out that that is unnecessary? This state of affairs has allowed an increasing split to develop over the years between the standard model and actual clinical practice within the field.

All the above pillars of the standard model range from wobbly to obviously misguided, in the experience of most of the clinicians active in this field. Nevertheless, proponents of the standard model seem so sure of their ground that they are willing to raise ethical concerns about people who operate by different lights, which happens to be nearly everybody. In the perspective of the standard model, neurofeedback clinicians have been making too many claims, for example with respect to Bipolar Disorder, dementia, Parkinson’s, recovery from minimally conscious states, etc. The list is long. In each case, however, the original “discovery” was fortuitous, which then led to replications. Every case counts here because we are not looking at small effects. Statistics are not needed. In formal terms, the “number needed to treat” (NNT) to establish a difference between the test population and everyone else is one or close to it. It is then also within the reach of a clinical office to gather evidence on the consistency with which such results are obtained. By the time anyone gets interested in mounting a formal study, the juice has already gone out of the issue. Some graduate student can get a Ph.D. doing a study, but probing minds will be on to other issues. In this fashion, a large split has opened up between the “standard model” proponents and the larger clinical community.

The only way this state of affairs could continue over such a long period of time (two decades by now) is that the proponents of the standard model were inclined to dismiss what they were hearing from the clinicians, particularly when that was in discord with their standard assumptions. Of course this turns the obligations of scientists upside down. Data should always be accorded primacy over models; in fact, data contradicting the model deserves special attentions; and even isolated anecdotes should be attended to. In the event, this has allowed the experiment of nature to proceed as well as it has, because the two sides have had essentially nothing to do with one another. Clinicians have spoken, but they have not been heard. On the other side, clinicians have been denounced, but they have not paid any attention.

It is now time to call for an accounting. The more reality-driven perspective turns out to be that of the clinicians, and their approaches have shown themselves to be superior in all respects to the approaches that follow from the standard model. All the outrageous claims that have been made over the years are turning out to be true. Every claim made by the clinicians is being supported by yet other clinicians using different methods. On the other side, the specificity being claimed for the “scientific” approach to neurofeedback, either in terms of protocol going in or in terms of final results coming out, is simply not there.

Dynamic Neuromodulation
The basic organizing principle that underlies what we term 'dynamic training' was enunciated succinctly by Walter Freeman: Brain structure and function are scale-free.How, then, may we think collectively about the emerging “clinically-based” model of neurofeedback, when in fact it consists of so many disparate approaches? The basic organizing principle that underlies what we term “dynamic training” was enunciated succinctly by Walter Freeman: “Brain structure and function are scale-free.” That basically means brain dynamics are not containable within a classic Gaussian framework. That in turn not only undermines the standard model in all of its generalities and particulars but it changes all the rules in feedback as well.

The second principle is that the brain places the greatest constraints on how it organizes itself spatially in the moment, whereas it places very few constraints on how it organizes itself over time. If we are to interact with the brain, it is more salient to address it in those respects which matter most to its functional integrity: the spatial map (Freeman).

The third principle is that cerebral architecture is organized as a “small-world” network (Barabasi), characterized by multi-stage hierarchical structure that exhibits small-world character at every stage. The result is a cerebrum that is functionally highly integrated, and can therefore be appealed to through feedback at any point and at any frequency. All scalp locations and all frequencies are not equal, however, and matters ultimately have to be resolved empirically.

The fourth principle is that long-distance communication via the white matter imposes tight timing constraints that must be met throughout the cerebrum. The brain will resist any interference in the regulation of its own timing because these matters are so crucial.

The fifth principle is that we should understand brain function in terms of action-perception cycles in which the brain’s agency is the first consideration, not the last. In a construction that goes back to Merleau-Ponty, perception is seen as the outcome of a preceding action. In nearly all of historical neuropsychology, perception has always been the starting point of the discussion. When we alter our perspective to place action first, the case for dynamic feedback immediately follows, without any subsidiary assumptions or hypotheses. The brain acts into the world, and the feedback signal is the result. According to Walter Freeman: “The action-perception cycle of behavior operates at all levels simultaneously.” Hence we can interact with it at any level.

What follows from this may be enshrined as the sixth principle: Causality cannot be thought of as having a singular starting point. It is always in process. Walter Freeman talks in terms of a circular causality: “In this view, the experimenter trains the subject to cooperate through the use of positive and negative reinforcements, thereby inducing a state of expectancy and search for a stimulus, as it is conceived by the subject.”

If we proceed from these principles, what are the implications for feedback? We recognize, first of all, that we are promoting function rather than expunging dysfunction. We recognize that we should proceed with a systems perspective on brain function. The hierarchical organization implies that even as we interact with particular subsystems, we are engaging the whole system. And when it comes to the tactics of neurofeedback, we have to appeal to the brain in terms that it most cares about, which is the organization of its instantaneous activity in its temporal and spatial aspects.

The most dynamic variable that reflects the brain’s regulatory activity in the domain of timing and frequency is the phase of the EEG. The phase governs Freeman’s spatial map. If we are to interact with it efficiently, we do so with bipolar placement that yields the instantaneous relative phase directly. This is what the brain cares about most and therefore regulates most assiduously. On the other hand, the phase does not persist. So we must track the relative phase with the highest available feedback dynamics. (That is why we have always emphasized the most rapid response of which the feedback loop is capable—in our development of the NeuroCybernetics originally; in our work with BrainMaster and BioExplorer, and now in our work with Cygnet.)

The relative phase is highly dynamic, so training criteria should be adaptive. Training according to fixed criteria would be like sailing by last month’s weather report. And if we are training according to the model of action-perception cycles and of circular causality, then feedback must be continuous, not episodic. We need to maximize the brain’s ability to “relate” to the ongoing feedback signal, and to discern the relevance of that signal to its own activities. That is why we introduced updating of thresholds in the NeuroCybernetics early on; went with faster updating in Brainmaster; adopted continuous updating in BioExplorer, and have moved to full dynamic thresholding in Cygnet. These considerations render the standard operant conditioning model inoperative for this approach to training. The original attention to discrete rewards becomes one of attending to an ongoing process. The operant conditioning model is not wrong. It clearly does describe some kinds of neurofeedback. It is merely being supplanted with something more inclusive. More to the point, the operant conditioning model cannot be used to constrain how neurofeedback is to be done.

So there we are. Training is most effectively done with continuous feedback on the brain’s management of instantaneous phase relationships. The brain will be challenged to react, and as a result the brain will quite naturally strengthen in its self-regulatory capacities. This process is not contingent on the proper goals having been dialed in, but rather depends only on the existence of the provocation itself. It turns out that Lester Fehmi appreciated the centrality of phase ever since he began his work in alpha reinforcement some forty years ago. It took others of us longer to see it. Alpha synchrony training is an obvious appeal to relative phase, and alpha amplitude training is just a more pedestrian version of the same thing. Alpha amplitude reflects local phase consistency. By extension, all of the standard SMR/beta amplitude training can be reinterpreted in terms of the phase-based model. The more one moved toward dynamic training with bipolar placement, the more effective the work became. The effects were so strong that individualization of training parameters became necessary. Tailoring the response to the individual was the final piece of the dynamic training model.

Corroboration of this model comes from an unusual source. Stimulation-based systems all operate according to the same principles. Here the phase of the stimulation signal sets up a discord with the instantaneous phase of the EEG, to which the brain reacts. In the case of ordinary AVS systems, as well as LENS and ROSHI, only the instantaneous relative phase matters in provoking the brain’s reaction. History does not come into it at all. Goals or thresholds don’t come into it. Diagnosis is also irrelevant. And the QEEG is entirely out of the picture. Only the process of challenging the brain subtly out of its existing state matters.

We have independent corroboration, then, of the basic premises of dynamic neurofeedback from the stimulation side of the house. So neurofeedback is finally being liberated from the constricting standard model, one that was not around to beset the stimulation technologies and hinder their progress. The response of the brain itself is the best guide to training. The right protocol does not have to be divined in advance by one means or another. Every nervous system has much in common with every other, giving us a general “Ansatz” based on consideration of functional neuroanatomy. But every nervous system also responds uniquely to neurofeedback, and so the training should be adaptive on all relevant time scales and with respect to all reward frequencies.

A Synthesis, and a Reflection
In truth, a number of us have been trying to maintain fealty to the standard model over the years, and have tried to paddle along with the mother ship. But the clinical realities have pulled us toward something quite different. Eventually the time is reached where the sinews of connectivity to the standard model are over-stretched and they snap. We have reached that moment. The old description no longer fits, and we need a new one. We must also acknowledge that those who still hold to the standard model don’t necessarily do so because they believe in it. They simply see it as a necessary step to gain mainstream acceptance. This is a vain hope, I fear. The mainstream isn’t going to suddenly turn friendly. The clash of interests remains. Eventually, the new paradigm will just take over, whether they like it or not.

Valdeane Brown was the first to abandon the standard model wholesale and to adopt the new thinking in the development of his system. His entire approach follows from the recognition that state transitions in the brain are subject to scale-free dynamics. The attention of the brain is called to its occasional large and inappropriate excursions in state space, and the brain learns to alter its behavior. The only action item in the whole approach is to monitor the state of the system in the moment, and to compare it to its immdiate past history. Large excursions in EEG “behavior” are flagged. There is no external reference point to restrict what is to be done, nor is anything specified with regard to how the brain should behave. In fact, there is no clinical decision to be made at all with respect to the program itself.

In the case of Val Brown, the transition to new thinking was explicit. It was also explicit in the case of Chuck Davis and the ROSHI. Here we have yet another system that requires essentially no clinical decision-making. It involves a generic challenge to the brain based on a minimal set of assumptions. The challenge lies in the phase domain, but it is distributed over the EEG frequency spectrum and to an extent over the scalp. By rapidly changing the reinforcement parameters, the brain never habituates to the challenge. These two systems represent opposites in one sense. Val Brown’s system does not utilize the instantaneous phase at all (it can’t, being based on Gabor transforms), and the ROSHI is based on nothing else. The LENS system is also entirely phase-based, but here the targeting is specifically related to what transpires in the EEG in the moment. The LENS is characterized by brief stimulations, which is consonant with the new model. If the stimulation is to be governed by the instantaneous EEG, then there would be no point in extending the same stimulation over the longer term because the brain would be off doing something else in any event.

In other cases, such as our own, the new model is only implicit rather than explicit. After all, training according to the standard model is also accommodated in our instrumentation. That’s where we started. So the evolution in thinking that has taken place could literally happen under cover, and the issues could remain obscured. The historical training protocols are merely subject to some subtle re-interpretation. And now that we have finally adopted the model in which the phase of the EEG is seen as being of paramount interest, we have become engaged with activity in the infra-low frequency region where that proposition is no longer self-evident. There will, no doubt, be another newsletter on this topic down the line…

The traditional approach to neurofeedback has in fact addressed itself both to inappropriate state transitions through the inhibit strategy, and to the organization of phase through the reward strategy. (Amplitude is here seen as reflecting local phase relationships.) So very little actually needs to change in order to accommodate the new systems-level thinking. The changes that were required in our clinical approach were minor. We moved from referential training either to bipolar placement or to synchrony training, depending on the objective, to enhance sensitivity to phase. We incorporated adaptive thresholding so that the information provided back to the brain remains most relevant to the present moment. The training is now adaptive to the person (in terms of reward frequency) on the basis of his or her response. And one chooses placement on the basis of what is known about functional neuroanatomy. The clinician has minimal duty with regard to operation of the instrumentation, and maximal responsibility with regard to monitoring the status of the client—because changes can happen so quickly.

These deviations in our approach from the traditional training are all minor, in one sense. But their significance was not missed. Proponents of the standard model raised objections at every point along this trajectory—with the shift to bipolar placement; with the adoption of inter-hemispheric placement; and with the migration to the low and the infra-low frequencies. By now, of course, this approach has been adequately validated across the board by thousands of clinicians in a number of countries and for a whole host of clinical conditions. What mainly needs to happen now is the shedding of the mythology surrounding the standard model, the necessity to act as if it had exclusive validity even when we know better. The rejection of neurofeedback by the mainstream is being mirrored in our field by resistance to the dynamic training model, and this appears to be for entirely strategic and political rather than for scientific reasons.

There remains a key distinction between the standard neurofeedback training and the stimulation modes of neuro-regulation. The standard training persists over the longer term with the same reinforcement parameters (i.e., reward frequency). This can be very effective in shifting the state of the person with respect to arousal, affect, attention, autonomic balance, etc. There are many conditions where this is in fact the highest priority clinically: Anxiety, depression, the Autism spectrum, Attachment Disorder, and PTSD. It even helps the neurofeedback process itself to hold the person in the most calm and alert state of which the nervous system is capable at that moment. Increasingly I see our kind of training as context-setting for whatever needs to get done, be that other kinds of neurotherapy, or even psychotherapy! In the hierarchy of needs from the perspective of neurophysiology, leading the person to the most controlled, well-regulated state of the nervous system ranks at the top.

We end up with a process that is highly effective in moving the person in state-space, and in stabilizing the brain, but we find that the process can be very sensitive to training parameters in the reactive or unstable nervous systems that populate our office. It is apparent that this kind of training cannot be subjected to research designs that require fixed protocols or even blinded operation. It is also problematic that the consequences of mere state shifts (which may occur very early in training) cannot be separated from the longer-term learning process that one would like to document. We used to chart our progress with autism in terms of percent symptom reduction per session, but that index is no longer meaningful when both factors are in play. Finally, it is problematic for research purposes that this kind of training is no respecter of diagnostic boundaries or of diagnostic thresholds. Clearly research methods need to adapt to the realities of what is being researched, not the other way around.

When we combine all of the neurofeedback approaches in the field which do not adhere to the standard model, they in fact constitute the bulk of the field. This includes in addition to our own methods the work of Lester Fehmi, Jim Hardt, Eugene Peniston, Val Brown, Len Ochs, Chuck Davis, Jon Cowan, David Siever, Tom Budzynski and others. Several among them have cultivated active communities of clinicians. Collectively, their clinical success supports the dynamic training concept rather than the standard model. And if there is any reality at all outside of the standard model, then it loses its exclusivity. It is time that this disagreeable fact be acknowledged, and it is to be hoped that the window-dressing for the sake of accommodation to mainstream thinking will finally be abandoned. So, let the hazing stop, and let the genuine listening begin. In a clinically-driven field, the clinicians should be heard on equal terms. It is time for our own entry into the New Romantic Age of Science.


1. Reclaiming Cognition: The primacy of action, intention, and emotion, edited by Rafel Nunez and Walter Freeman
2. Neuromodulation Technologies: An attempt at classification, Siegfried Othmer, Chapter 1 in Introduction to Quantitative EEG and Neurofeedback, Tom and Helen Budzynski, James Evans, and Andrew Abarbanel, editors.

Siegfried Othmer, Ph.D.

11 Responses to “The Great Divergence”

  1. val brown says:

    An interesting summary but you have gotten at least one important point wrong. Using Adaptive Gabor transforms and does allow the calculation of instantaneous phase so the complete truth is that NeuroCARE was — and still is — completely out in front of the rest of the field.

    • I’m not surprised at your response. The fact is that whenever transforms are used, of whatever kind, integration over the conversion window is involved, so the calculation can never yield the “instantaneous phase,” but rather some measure that refers to the entire conversion window.

      Now it is true that any measurement method whatsoever has to sample the system over some period of time, but it is beyond controversy that filtering approaches can get you closer to real time, to instantaneous phase, than transform-based methods.

      With our system, an EEG event registers at the feedback screen within 250 milliseconds. That cannot be done with transforms, and certainly not with Gabor transforms, given the usual set of parameters used in EEG work. The relative phase at a frequency of interest—in the beta band, for example— can be tracked essentially at the sampling rate, and updated at the video refresh rate, which is to say with exquisite fidelity, whereas with transforms you can track changes only at the conversion rate. Such relationships at beta frequencies would not be discernible with transforms with comparable fidelity.

      With the stimulation technologies like the ROSHI, there is no delay or loss of fidelity at all involved, since no measurement is being made in any event. No training instrument that depends on EEG measurements can match that.

      However things may be in principle, the fact is that your system has not made use of phase information for as long as I have been acquainted with it, at least in the manner that I have discussed it. When we talked years ago about your adding a Lissajous display to your system to yield the relative phase of the two channels at a given frequency, apparently that couldn’t be done with your software at the time. In any event, it wasn’t done.

      Your system is poles apart from the approaches used in LENS and ROSHI and Cygnet, where phase information is paramount. So even if phase somehow plays into your designs at some level which you haven’t shared, that is clearly not in any way comparable to the way LENS or ROSHI are using phase, nor to the way Cygnet is utilizing phase information. When it comes to phase-based training, you are simply not part of the conversation. Sorry.

  2. JERRY GLUCK says:

    A provocative set of ideas, but you got a number of things wrong. While the brain is an open system, this does not mean the gaussian distribution is inappropriate to its function or correction of function or dysfunction. You mix metaphors and conceps and come to distorted clinical directives. What follows is a bare summary. First, while Freeman said scale free your way of using it is a distortion and out of context. Second, QEEG not does as you put it traditional neurofeedback exclude older folks nor young children. IN fact the youngest child was a little older than an infant, a toddler whose therapist was Margaret Ayers. But then again I am not sure what you are calling young. A child has to be able to tolerate sitting at the computer and know the difference between on and off. Children’s ability to spend time in front of the computer can be extended and shaped when very short. Third, bipolar training probably does change phase, but it does so in an unpredictable manner without the assistance of a metric. Why not just directly train phase or coherence? Fourth, amplitude is not an expression of phase in the sense of timing or lag, but an expression of the number of neurons firing at the same time which is the result of thalamic cortical timing of the numbers of neurons recruited during phase shift, but the timing of the shift is different than the numbers recruited to be synchronous. Fifth, to suggest that “traditional” NFB practioners are left brain soley is absurd and totally distorts how these folks practice. Sixth, depression and anxiety and most comorbid disorders are clearly not excuded from treatment in ADHD in clincial practice. They are treated either sequentially or simultaneously depending on the situation. Children and adults are selected for training of course on the basis of EEG (wouldn’t a hematologist select treatment on the basis of blood results?) BUT ALSO on the basis of symptoms and behavior. This is probably long enough, but what makes this particularly vexing is your continued distortions and thereby distorted conclusions to direct clinical practice embedded in the context of some sound theory and a complex writing style. Clinicians should be cautious in applying these notions without also consulting what you would call traditional science.

    • Gerry:

      I can’t deal with the issue of the Gaussian distribution in a paragraph. I’m sure to get back to that topic in subsequent newsletters.
      But to some of the other issues:

      Of course the QEEG model says nothing about whether the young or the old are trainable. But you remember perfectly well that Lubar insisted for years that young children could not do neurofeedback, and he maintained this position against all of the cumulative evidence from Ayers and ourselves and others. That puts theory above data and renders it impervious to data.

      You write: “A child has to be able to tolerate sitting at the computer and know the difference between on and off.”

      I disagree, on the basis of plenty of counter-evidence. Some of our very young children (near-drowning victims, etc.) are so impaired that they don’t have a clue what is going on. But their brains train nicely anyway. One of our clinicians just had a baby, and she trained her infant when she became colicky. These facts compel a reappraisal of what is actually going on in feedback, and that’s what I am attempting to do.

      You write: “….bipolar training probably does change phase, but it does so in an unpredictable manner without the assistance of a metric. Why not just directly train phase or coherence?”

      Why use a metric when you don’t have to, particularly when it comes at great expense? Bipolar training has a built-in bias toward desynchronization. But it helps even those where desynchronization is not the objective. That means a metric is usually superfluous. The brain moves toward better self-regulation even if we don’t steer it that way. That is the implication of all our work, as well as that of the stimulation technologies. Of course sometimes a measure is useful, and then you go after it. I am not arguing against that. I object to the exclusivity being claimed for a single approach.

      You write: “…to suggest that “traditional” NFB practioners are left brain soley is absurd and totally distorts how these folks practice.”

      I was referring to the standard-setting researchers here. I’m sure there are plenty of neurofeedback practitioners with a functioning right hemisphere who are trying to master the standard model because they have been told that they must. I was not trying to disparage the left-brain perspective by any means (I tend that way myself), but rather to make room for the right-hemisphere perspective at the table.

      You write: “…depression and anxiety and most comorbid disorders are clearly not excuded from treatment in ADHD in clincial practice.”

      Again, I am sure you recall the many years over which Lubar maintained that anxiety and depression needed to be ruled out before an ADHD child would be accepted into neurofeedback training in his facility. This went against all of the accumulating clinical evidence regarding neurofeedback for anxiety and depression in children, so we have here yet another instance of a model impervious to data. Additionally, however, it illustrates compartmentalized thinking that in the emerging systems perspective will increasingly be seen as quaint. We learned how to work with depression while dealing with ADHD, so they are not worlds apart.

      You write: “Children and adults are selected for training of course on the basis of EEG (wouldn’t a hematologist select treatment on the basis of blood results?)”

      You consider the EEG as somehow comparable to a blood test in hematology? I think we’re getting to the crux of the problem. Would a neurologist rule out epilepsy because the EEG is clean? Of course not. Would we refrain from training an epileptic child who had a clean EEG? Of course not. We are training the system; the EEG is just the means by which we garner the attentions of the brain to effect its functional renormalization. The EEG does not also have to manifest the pathology.

  3. val brown says:


    Actually again you’ve gotten it wrong but, in one small way, you’ve also said it exactly right. You say: ” That cannot be done with transforms, and certainly not with Gabor transforms, given the usual set of parameters used in EEG work.” The operative phrase is “the usual set of parameters used in EEG work”. We don’t use “the usual set of paramaters” and that’s also a central part of what sets our system apart from the rest of the field. It IS unique and, despite many claims to the contrary, has not ever been copied or even emulated, although a number of others have tried. You also say that: “When we talked years ago about your adding a Lissajous display to your system to yield the relative phase of the two channels at a given frequency, apparently that couldn’t be done with your software at the time. In any event, it wasn’t done.” The truth is that we do have a Lissajous display, it just works differently than what you wanted and what “wasn’t done” was what you wanted, per your spec. Relative phase IS displayed in NeuroCARE and has been for years. But in some ways these points, however important, miss a central point. Our approach works and works as well across the board as virtually any other approach and that is shown in our user survey. And it does that without tracking “symptoms” (in order to “change the parameters”) and it does that without using QEEG or other assessment and it does that with trainers who have essentially no background in neuroscience or other related fields. The software really does handle the process and it does that in ways that no other software can. Now that’s not to say that other approaches aren’t any good — they certainly are. But it is to say that the claims that neurofeedback is “specific” (by site, “diagnosis”, “frequency”, etc) are clearly incomplete at best because out approach is 100% diagnostically agnostic, uses the same central sites all the time and trains all relevant time-frequency events in the currently accessible range. I appreciate your comments — I really do — I would only ask that the more complete story be told.

  4. Val—
    I don’t know why you can’t leave well enough alone. I acknowledged your system for its virtues—effecting renormalization of brain function without regard to diagnosis, and without protocol particulars such as placement and frequency selection. It is highly suitable for an undoubtedly large group of clinicians who want to have the benefit of neurofeedback for their clients without necessarily making it their career. But when you say that your approach “works as well across the board as virtually any other approach and that is shown by user survey,” you overreach. We’re in a field where people are incredibly sensitive to what they see as over-claiming, and this seems like a case in point.

    There are very few occasions to make fair cross-platform comparisons. One of the few I’ve encountered was the Swiss group of parents of autistic children that I just had the chance to address in Zurich. They are no longer recommending just neurofeedback to their members; they are recommending the Othmer Method specifically. They have found it to be superior for autism and Asperger’s. Now your system is popular in Switzerland. So in this comparison, it is undoubtedly your approach, along with the more conventional methods, that they are no longer recommending.

    So something is indeed gained by adding the phase-based training. If someone doesn’t want to go to the trouble of acquiring the skill to manage all this, that’s understandable. But this is different from arguing that there is little or nothing to be gained. While formally claiming nothing, you implicitly subsume all of the claims made for the other systems.

    Now to the specifics: For your Gabor transforms you have no choice but to use a set of parameters suitable for EEG work. Whatever conversion window you choose to represent EEG information will give you two ineluctable problems: 1) Any phase information derived from such transforms will always be stale with respect to matters of import to the brain, and hence largely irrelevant to phase-based training; and 2) any such window will be too short to accommodate the infra-low frequencies, which is where we now do most our work. (The work in the infra-low range of frequencies accounts, almost in its entirety, for our ostensibly superior outcomes for the autism spectrum.)

    With regard to the Lissajous loop, if your system had been able to yield the plot I was suggesting, it would have been a straight-forward matter to provide it (once you had the display graphics done). In BioExplorer, which is oriented toward both frequency-based and transform-based analysis, it is a matter of ten minutes to set that up. Your system could not provide that display, quite clearly, and that makes my point. Your system was not then, and is probably not now, capable of supporting phase-based training as it sits. So you have to fall back on your position that this is somehow superfluous. Our experience, as well as that of the ROSHI and LENS and AVS communities, says otherwise. It is becoming quite clear by now that no single system covers all the bases if one wants to exploit the potential of neurofeedback fully.

  5. val brown says:

    Sieg, you really don’t know what you’re talking about when it comes to our system. But perhaps you misunderstood what I wrote about not including the Lissajous Display as YOU wanted it done. That was a deliberate choice, not a limitation in our system. Similarly, it was a deliberate choice to not include traditional measures, even though they were also included in the original spec.

    We don’t need them. They don’t add to the results — as reported by our users. And the same is true for what you’re now calling “phase-based” training.

    The plain truth is that our system was the first to offer training based on filtering that extended down to 0.001 Hz as a lower frequency. In the old days we used FIR Hybrid median filtering, when computer resources to support real-time Adaptive Gabor transform weren’t widely available. But we’ve now been using the Adaptive Gabor approach for years, even though all of the other options are there as well.

    Now let’s also be clear. I’m not saying to ONLY do neurofeedback. I’ve never said that. What I am saying — and what I have been saying for years — is that there is really no better way to do neurofeedback than what is available in our system. And what’s remarkable is how the comments from other systems are still the same as they have been over the years. There is always a comment about some “limitation” in our system and about some “new discovery” in another system, many of which Sue and I first presented on and, when we presented, we were told it wasn’t possible, wouldn’t work…

    Look over the old posts about low frequency training and see the comments others made back then about how NO system could do low frequency training, unless it was SCP. We were doing it then, and we’re still doing it now, regardless of how much it seems impossible.

    Now I understand that this is your platform and it probably wasn’t a good idea for me to jump in here but there were a few points that needed to be addressed, esp when it comes to how well our comprehensive, adaptive targeting process works — and how upsetting it still is to the rest of the field that people can learn to use our system within about an hour and provide robust neurofeedback training easily, safely and with confidence.

    Now I am happy to discuss the particulars of these issues in more detail if that’s really wanted here so let me know. Otherwise, do stay well and keep moving on!

  6. Very interesting article.

    Thanks for posting.

    I just finished my new book on neurofeedback,

    and I’d love to get you a free copy of the audio book (two and a half hours).

    What email should I send a link to the book to?

    The name of the book is:

    “Neurofeedback: Transforming Your Life With Brain Biofeedback”

    It will be up on in January.

    I work in Orange County, CA.

  7. Rosa says:

    I´d like to know how Lissajous display could be interpreted.

    Thanks in advance,

  8. The Lissajous loop, just to explain, is a graphical display of the phase relationship between two sinusoidal signals. In general the display looks like an ellipse that is constantly changing, but when both signals approach a common phase, then the elliptical signal collapses into a line.

    Our intention is to have this be experienced rather than interpreted. The signal is typically very dynamic, and so it does not lend itself to easy interpretation. But the brain does engage with it, and respond to it, which is more the point.

Leave a Reply