Empirical Evaluation of Inter-Hemispheric Training

by Siegfried Othmer | October 20th, 2005

To accompany Sue’s description of the historical development of our clinical approach at the 2005 ISNR meeting, I presented our current data obtained with inter-hemispheric training. These results were compiled by John Putman, and were also just published in the Journal of Neurotherapy (Volume 9 (1), pp 37-52). The published data were restricted to those cases where inter-hemispheric placement was used exclusively. This necessarily left out quite a number of cases where the inter-hemispheric protocols predominated, but where we also employed more traditional lateralized trainings to address particular concerns. We see the inter-hemispheric training possibilities simply as an expansion of the “space” in which we can operate to good effect. There is no reason to abandon the techniques carefully worked out in prior years.

In the following, I would therefore like to “expand the space” of our considerations as well. What is the best that we currently know how to do, irrespective of protocol particulars, and how does that compare with the results we have obtained in the past? We still operate under the rubric of inter-hemispheric training because the latter has come to dominate our clinical work. In this comparison, it may be best to focus on impulsivity as an index. Firstly, we have good data on this from the TOVA measurements that we have been doing since about 1992. Out of all the TOVA data that could be surveyed, we want to focus on impulsivity particularly because it is most likely to reflect the state of disregulation of the system. A certain cerebral functionality has to exist for a person to show up as impulsive in the first place. So one might readily postulate that a brain sufficiently functional to show up as impulsive might also be capable of operating in the normal range. In other words, we can anticipate that nearly universal success might be achievable in the regulation of impulsivity through neurofeedback. The same cannot be said about the inattention measure in the TOVA, where organicity could be much more of an issue. Results on the attention subtest might therefore depend much more strongly on client characteristics about which we know very little at the outset.

In Figure 1 we show the impulsivity data that appeared in the recent publication. In Figure 2, we show the larger data set that also includes cases where some lateralized training was done along with the inter-hemispheric placements. Both sets of data refer to nominally twenty-session retest data, and hence do not represent the final end-point of training in most of these cases. This is done simply because the CPT normalization tends to occur early in training, and the issues addressed in subsequent trainings often target other issues than those captured in the CPT data. Also, the twenty-session milestone gives us a larger “n” to work with. Finally, there is the issue of comparability with earlier data, which also used a 20-session retest for comparison.



In Figure 3 we show early data that were compiled by David Kaiser on our prior lateralized trainings (“C3beta” plus “C4 SMR”). These data refer to work done some ten years ago. The data are rank-ordered by level of initial deficit in standard score. In Figure 4 these same data have been grouped into bins of different range in deficit. Significantly, the group in greatest deficit, at least four standard deviations below norms, collectively shows progress by three standard deviations, which puts the cohort well into the normal range of functioning. Observations like these gave us the impetus originally to set an expectation that impulsivity should be normalizable in nearly all of our clients.



Comparisons between the historical and contemporary data are shown in Figure 5. At all levels of initial deficit, the modern data show greater improvements than the earlier. This comparison should be taken as an indication only, and we have not undertaken to do any statistical analysis on the comparisons. A formal statistical analysis would be inappropriate because the client groups involved were in fact radically different. In the early days we tended to see more standard-issue ADHD kids who would respond readily to the standard approaches of the day. Currently we see clients who are more seriously impaired and more broadly distributed across the age range. The fact that the new techniques show numerical superiority despite this disparity in groups is a cogent confirmation of the relative advantage of the new methods.

There is an additional factor in play, and that involves the built-in bias of looking only at those clients who in fact stayed the course to the twenty-session retest. In the earlier day, the more complicated clients would be more likely to attrition out of the program. Currently, we are bringing more of these difficult clients to completion. One reason for this is that we have moved back to a clinical model of “one-on-one” training with a clinician, which tends to hold people better. In the earlier day, reliance on more standard protocols allowed us to implement a model where one clinician could manage several training stations at the same time. (Under that model, we had one month in which Sue’s office ran 586 NF sessions with only two additional staff people.)

The second reason is directly traceable to the adoption of the inter-hemispheric protocols. The process of optimization of reward conditions is necessarily much more interactive than the traditional approach, and the person is usually aware of state shifts attributable to the reinforcement, even in the first session. Such awareness of state shifts serves for us the same incidental function that side effects contribute to medication compliance: they convince people that something real is happening, and that they need to be engaged with the process and take it seriously. So we get much better “buy-in” by the client right out of the gate.

Showing Figure 5 also gives me the opportunity to comment that the figure of merit of change in TOVA score per session is seen to be strongly dependent on the level of initial deficit. That being the case, it is not possible to give a single such figure of merit for an entire clinical population that is scattered all across the range in deficits. Such a global figure of merit would say more about the properties of the population being trained than about relative efficacy. Nevertheless, Hershel Toomim attempted such a comparison, and came up with the figure of merit of 0.5 TOVA points per session for EEG Neurofeedback, and about 1.0 TOVA points per session for HEG NF. Nearly all the data points in Figure 5, both early and recent, exceed the 0.5 points/session that Hershel calculated from the early neurofeedback studies, and most of them exceed the level of 1.0 points/session that Hershel calculated for HEG. Future comparisons of this kind will have to match levels of initial deficit, as well as matching test-retest interval in the number of sessions, in order to yield meaningful comparisons. Meanwhile, it should be noted that in the presence of the largest deficits we also make the most rapid gains, up to 2.8 TOVA points per session. That’s nearly six times larger than what Hershel calculated for the whole training cohorts that he was looking at.


Finally, it is of interest to show the distribution of reward frequencies that were found with these clients. This is shown in Figure 6, in which are plotted the values employed at the end of the twenty-session sequence, by which point they have typically stabilized. Moreover, the choice of reward frequency will have been “confirmed” by virtue of consistency observed over the entire training history.


The findings are highly illuminating. There appears to be a trend toward reward frequencies at or below the traditional “SMR” band of 12-15 Hz, but within that broad frequency range there is no predominant pattern. It will at some point become clear just what puts people into certain frequency domains for their best training, but we are not yet seeing this clearly. We do know that the most highly over-aroused folks, in our standard classification, train at the lowest frequencies.

Those who have been following this field a long time will be aware that originally (1988) we had adopted Margaret Ayers’ approach of training only the “beta” band of 15-18 Hz. When the shift is made from left-side training at C3-T3 to inter-hemispheric training at either C3-C4 or T3-T4, there appears to be a downward shift in optimum reward frequency that is reminiscent of what we found with right-side training. But inter-hemispheric training appears to be much more sensitive to the particulars of the reward frequency than lateralized training, and thus it was necessary for us to expand our reach all the way to the lowest frequencies accessible to us in order to optimize everyone’s response.

If there is a distinct category here that should draw our attention, then it is those who train at the very lowest frequencies of 0-3 Hz. This comes very close to the regime already occupied by the Birbaumer group with their training of slow cortical potentials. It could be that what is in store for us is a “unification” of the frequency-based training with the “time-based” training of Birbaumer & Co. There appears to be a clear benefit in terms of training efficacy, or at least in training efficiency, if certain folks are challenged to organize their delta band (in the frequency-domain picture) or their slow cortical potentials (in the time-based picture).

Secondly what might draw our attention is that we can train perfectly well with reinforcements in the theta band, even in cases where theta amplitudes are substantially elevated. This is simply an observed fact, and it cannot be wished away. Rather, these findings mandate a reappraisal of what such continuous reinforcement training is all about when deployed with bipolar placements, both inter-hemispheric and otherwise. This topic will be dealt with in a subsequent newsletter.

The publication of our data on inter-hemispheric training in the JNT was followed in the next issue by Sue’s article in the Clinical Corner on how this technique is currently deployed (Volume 9(2), pp 87-96). Complementing all of this was Sue’s presentation at the ISNR of the historical development of these protocols. This presentation is available from the ISNR as a DVD, and will soon be available from EEGshop.com as well. Sue’s Power Point presentation has also been posted on our Bulletin Board (for members only). Cory Hammond, serving as editor of the Clinical Corner, had invited Sue to write her article a year and a half ago, and he introduced the piece in a fashion that bears further comment.

Cory Hammond’s Commentary

“In this issue [Volume 9(2) of the Journal of Neurotherapy], Clinical Corner examines interhemispheric neurofeedback training.” Thus starts Cory’s introduction, setting the stage. But then he goes on to say, “Margaret Ayers, Chuck Davis, Valdeane Brown, and Susan Othmer are some of the pioneers in the area of simultaneously training the two hemispheres, often at homologous sites and each utilize different approaches.” Perhaps the intent here is to embed the interhemispheric training in some larger context, or else it may be a matter of lumping the principal mavericks of the field efficiently together in one sentence. But then Cory goes on to say, “Margaret Ayers … is perhaps the first clinician to have used bipolar interhemispheric training.” Oh, really?

Lest I be misunderstood in the following, let me hasten to say in preface that I have always supported Cory in his continuing attempt to make sure that the work of Margaret Ayers gets recognition within the field. Ayers was clearly the first person to comprehend the breadth of the clinical impact of “SMR/beta” training, and she was essentially alone in this recognition for a long period of time–more than a decade, in fact. This deserves to be acknowledged, and the Society has done its duty in that regard. Ayers also pioneered the technique of dynamic (i.e. analog) reward-training which has essentially taken over in the field at large. The capability was obviously already there in the instrumentation that had been made available to her from the prior developments in Sterman’s lab, but she shifted attention to that aspect of training in a manner that had not been done before. By training dynamically, she achieved for the first time what is probably the essence of the process, the facilitation of a recognition by the brain of a correlation with the feedback signal. The result was a major step forward in training efficiency.

The significance of this advance might well be missed were it not for the fact that Sterman is now making a very public case for a restoration of the more traditional operant conditioning paradigm, with its emphasis on discrete rewards followed by refractory periods for the consolidation of learning. His giving that talk could not give stronger testimony to the proposition that the field has largely moved away from his earlier model. It appears that Sterman may never have abandoned that model at all, but rather just found new evidence to support it in the contemporary research on event-related desynchronization. Lubar is clearly thinking along the same lines in the way he talks about limiting the reward incidence for adults and children. So it was in fact Ayers who was responsible for taking the field in an entirely different direction.

By the simple expedient of making the discrete rewards (the beeps) plentiful rather than sparse, the training shifted to where the beep became the expectation and the dropout of the beep became the novelty. This had more the aspect of an inhibit-based approach, and it also shifted attention of the subject to whatever else was going on, namely the fluctuating little green light. Ayers’ major clinical findings mostly came relatively early in her work, and it bears mention that all of this happened with no more than a fluctuating green incandescent bulb and a repetitive beep. Did Ayers make this shift toward high reward incidence consciously and deliberately, or was her “threshold-setting behavior” shaped by feedback at yet another level? No matter. Adaptation ensued, and the field took off. What in Sterman’s hands had been the training of particular altered states, first in the cats and then in people, became a more general state regulation technique in the hands of Ayers.

All that said, the proper codification of the history of this field, particularly through our journals, should not go beyond the facts into the realm of fiction. Hammond cites as his basis Ayers’ chapter in Evans and Abarbanel. Now as I survey that chapter, I find no mention of interhemispheric training at all. This chapter was published in 1999. That was four years after Douglas Quirk published his paper in the JNT on neurofeedback for violent offenders, which used inter-hemispheric training exclusively. Quirk was reporting on more than 25 years of work, so he might well be considered the person whose work most deserves to be highlighted on this occasion in its pioneering role.

As for the original source of the protocol, things must go all the way back to Barry Sterman, although he does not wish to take the credit. Either that or Douglas Quirk was a bad note taker when he heard Barry lecture on the subject back in 1970. When Quirk began his work, he surely did not know enough about neurofeedback to be inventive. The fact that he stuck to a single approach throughout his career also suggests that he was not inclined to be inventive in this particular regard. He clearly took what he learned from Sterman originally and found that to be adequately effective for his purposes throughout his remaining career.

And we all know George von Hilsheimer. He has been beating the drum for the basic interhemispheric protocol at C3-C4 for a number of years, and has claimed to have used what he calls the “Quirk-Sterman” protocol ever since he and Quirk started collaborating. Now it is true that George’s work is not represented well in the literature, but then neither is Margaret Ayers’. If one goes beyond the written literature to complete the historical picture, should George’s role not be recognized as well?

Finally, there is our own first venture into inter-hemispheric training, which came about in a rather strange way back in about 1997. We wanted to address the category we identified as the “instabilities” with bilateral bipolar training using a combination of T3-Fp1 and T4-Fp2. Unfortunately the latter placement led to our experiencing emotional melt-downs and disinhibitions. Later this was traced to our having used reinforcement frequencies that were too high, but at the time we simply defaulted to the combination of T3-Fp1 and T4-Fp1, which worked fine. We gave some thought at the time to what the implications might be of our training across the midline, but we had little to go on. As has been typically the case all along, the clinical results grabbed us by the nose and led us forward.

One might also mention the work of Jonathan Walker in this connection. He has been targeting coherence anomalies, many of them inter-hemispheric, for quite a number of years with both up- and down-training. And that work is also published.

We can even bring matters up to the present. According to those in attendance, Ayers’ recent workshop handout mentions only a single interhemispheric montage, F3-F4, in the course of what is mainly a presentation on the lateralized training approaches that she has used all along. Cory reports in more detail that currently Ayers “…may train at F3-F4, O1-O2, P3-P4, and also T3-T4 when there are not problems with migraines, seizures, or a history of head injuries.” It is for this sentence that the Ayers chapter is cited, but Cory writes in the present tense, so presumably this is intended to reflect current practice in Ayers’ office.

This quotation actually frames the issue nicely. Ayers does not use T3-T4 training precisely in those situations for which Sue Othmer finds it most useful and most directly applicable. We were first driven to the protocol for the instabilities, and that largely sustains our continuing interest. The breakthrough issue is therefore not the use of inter-hemispheric training per se, or two-site simultaneous training, or homologous site training, or even bipolar training of any kind, but rather the use of reinforcement bands in the hitherto “forbidden” regions below the alpha regime with a bipolar montage.

For some reason, Cory felt the need to submerge this crucial issue in his introduction. And perhaps for the same reason, the Journal editors saw fit to surround Sue’s piece with all sorts of irrelevancies as if to inoculate the audience and buffer the Journal against criticism. Sue’s article has been sitting around at the Journal for a year and a half, after all, awaiting publication. At a minimum, it would have been nice to see Sue’s contribution to the Clinical Corner as a companion piece in the prior issue that contained our clinical data. Instead we got two theoretical pieces on temporal lobe training. Did these sit around even longer, awaiting print? That is doubtful. It seems more likely somebody was treating Sue’s article as a hot potato.

Of course there has to be editorial latitude with regard to the pacing of publications. But when I put all of these elements together, I see an attempt to shine the light nearly everywhere except where it belongs–namely on the breakthrough that has in fact occurred. Let there be no doubt that the leading lights in the field knew immediately what they did not like about Sue’s approach: it was the use of low reinforcement frequencies first and foremost, and only secondarily it was the issue of interhemispheric placement.

The matter about which all the critics had an immediate visceral certainty was that it was just a bad idea to reinforce in the theta and delta bands. And they mounted the ramparts in rebuttal. For example, Gunkelman told everyone within earshot repeatedly that there was no literature support for the protocol. For him Douglas Quirk did not exist either. And by the time he rose up in opposition to this approach, Andrea Sime had published her article on inter-hemispheric training as well (for trigeminal neuralgia). The controversy finally rattled the new owners of EEG Spectrum International, imposing a stress that eventually culminated in the dissolution of our working relationship. Spectrum management has since back-tracked to the old combination of “C3-beta” plus “C4SMR” for their beginning training course.

Many of the clinicians were so spooked about reinforcing at low frequency that they have not been willing to do so to this day. With all this brutal fallout over a period of years, and continuing even to the present, the Journal gives the impression that this issue does not even exist. Lots of people are doing two-site training, and Sue Othmer is to be found somewhere among them. What else is new?

This controversy illuminates so many issues that are crucial to our field. One issue concerns the role of the “scientist-practitioner” in the field, and whether such people are truly welcome. Another concerns how controversies should be aired within our journals. If the Clinical Corner is to be something of a “free-speech” zone for ideas not yet fully accepted, then why all the publication delays? Why must controversies within science be marginalized, and treated as if they were an imperfection in the process? Finally, and most fundamentally, a question is raised about our basic fealty to data. Why are basic clinical data papers harassed for so long, relentlessly massaged until all the novelty is squeezed out of them? As far as I can tell, there is a clear bias in favor of clinical papers that conform to already established beliefs. That is not how things are supposed to work in science.

Finally, I must comment on the pervasive tactic to gain power and influence by mobilizing fear. In this, things seem to mirror our national politics. Perversely, it is the anti-claim that gets nearly universal respect, while positive claims are flogged and hazed to within an inch of their lives. If someone asserts that a particular protocol does harm, then that person immediately commands attention. The anti-claim is almost implicitly believed. The positive claim, on the other hand, is held in abeyance or even dismissed as being anecdotal. As we have pointed out previously, the shoe is precisely on the other foot. A positive clinical experience will lead naturally and organically to many replications. (This could even be considered another manifestation of Thorndike’s “Law of Effect.”) These replications collectively shape and refine the clinical appraisal of the new approach. Chances are that many such replications will have been observed before the observer ever goes public with the finding. The plural of anecdote is replication, and replication leads to hypothesis-testing, which in turn leads to scientific validity.

On the other hand, if a clinician has a bad experience or two with a particular approach, that may be enough to kill the appetite for that particular exploration. There is simply not going to be much opportunity for a replication. So, even a single bad experience may be enough to launch someone forth on a campaign of “Beware! Beware! Beyond there be Dragons.”

In this way we can understand Ayers’ reluctance to employ T3-T4 training for migraine, seizures, and TBI. The difficulty lies with her use of beta reinforcement. As Figure 6 illustrates, that band is too high for the vast majority of clients of any stripe. In recent years Ayers has insisted that the bulk of her approach relies on inhibit-based training, but there is just no way that pure inhibit-based training can lead to a general difficulty with the management of migraine and seizures using T3-T4, and even less with TBI. At worst it is innocuous; it cannot be problematic in the general case. The difficulty must therefore be traceable to the use of beta reinforcement, and this observation is sufficient proof in my mind that beta reinforcement still constitutes a major part of Ayers’ protocol to this day, as indeed it has always been in the past.

What Ayers may mean when she says that she doesn’t do beta up-training is that beta amplitudes are not raised in the process, which of course is something we have all recognized for a long time. When the EEG normalizes in the cases of TBI that she sees, the whole EEG calms down, including the beta band in which nominal “up-training” has taken place. So reinforcement of beta amplitudes in bipolar placement at C3-T3 does not train up beta amplitudes as a final outcome in the general case. Well, guess what? Reinforcement within the theta band using bipolar placements does not raise theta amplitudes either. It is time that this should be generally understood.

The upshot of all this, finally, is that the use of low-frequency reward-based training should be regarded as a settled issue, and we should move on to the next frontier. That frontier includes teasing out the mystery of why a single reward frequency in any individual is sufficient to achieve improved self-regulation across the board. Different symptoms tend not to require different reward frequencies at a particular site (although these frequencies will certainly be different at other sites). More and more, as our field matures, the question of how a “disregulated” person trains becomes separated from the issue of what symptoms he brings in the door. This has been true for some while with regard to QEEG-driven protocols. It has been true for Val Brown’s more dynamical approach. It is also been true for what has come to be called “symptom-driven” training. That is to say, the protocols remain fairly standard for nearly all problems of disregulation, as indeed they have been all along. The symptom tracking has to do with refinement and optimization and progress measurement rather than with determining the protocol going in.

A second conundrum is why the reward frequencies at different sites stand in a relatively fixed relationship to one another in nearly all individuals. Both of these empirical ‘facts’ are suggestive of a web of connectivity across the frequency domain that is far more intimate than we had any idea about. Such a web of connectivity in both the spatial and frequency domains in turn makes it reasonable that our clinical approaches may always remain a relatively small set. Further, such small sets of approaches can be variously deployed. There is just a variety of ways in which we can couple into the system and mobilize a renormalization. We should then expect a lot of generality in the effects of our training, in conjunction with some particularity related to the actual training site. Jointly the above rounds out our case for the “Disregulation Model” of psychopathologies, and for standard protocol-driven neurofeedback as the generalized “Self-Regulation Remedy.”

Leave a Reply