Infra-Low Frequency Neurofeedback

by Siegfried Othmer | October 31st, 2008

Over the last several years we have increasingly explored the low-frequency domain of the EEG for neurofeedback applications. Surprisingly to us all this has led us to the realm of infra-low frequency training, below the 0.5 Hz cutoff that is commonplace in EEG work. It is tempting to refer to this very low frequency neurofeedback as a breakthrough, but in fact this has simply been the latest iteration of a long progression from one milestone to the next in an evolution of our particular protocol-based approach. So it does not feel like a breakthrough to those who have been involved every step of the way. It only appears like a breakthrough because when considered on its own it seems like a complete negation of the standard assumptions of neurofeedback. Perhaps in consequence of that the method has already attracted the usual gallery of skeptics. The survival of this kind of skepticism at this late date reminds us that the typical understanding people have about neurofeedback is being grossly violated. A reappraisal of the usual assumptions is therefore in order.

First of all, at the low frequencies extending down to 0.01 Hz it is no longer clear whether we are just seeing EEG phenomenology or whether the signal is dominated by other physiological variables. Clinically of course it does not matter as long as the brain responds to our reinforcements, but it matters a great deal to our understanding. Secondly the issue has been raised as to whether we are seeing a valid signal at these low frequencies. It turns out that the signal we are tracking is not small at all. In fact, in typical EEG measurements the very low frequencies are eliminated by filtering just so they won’t get in the way of what we are trying to do. If the low frequencies had been left in the signal the early researchers would have been watching the EEG bouncing wildly up and down on the oscilloscope screen, which would have been quite a nuisance. The conventional EEG signal we have been looking at just sits on top of much larger undulations at the lower frequencies, much like foam on an ocean wave.

The fact that the observed signal is large is not quite enough, however, to dismiss the concern that it may be compromised by drift in electrode voltages—particularly of gold electrodes—and by other extraneous factors. Drift in the gold electrodes is indeed severe, but it settles down over time. And the use of other electrodes, such as sintered silver/silver-chloride, gets rid of the drift sufficiently to exhibit the “real” signal. The final proof that we are not looking at something artifactual is provided by the training itself. The effects of the low-frequency reinforcements can be quite prompt and quite dramatic (in terms of induced state shifts) in certain clients.

Then there is the issue of how one actually trains on such slow signals. All we can do with a signal centered at 0.01 Hz, where a single cycle takes 100 seconds to unfold on the screen, is to reinforce on the instantaneous signal. This is not new either, and that brings up yet another reason this training is not really a breakthrough. The group doing neurofeedback under Niels Birbaumer in Germany has been doing this kind of work for decades. The difference is that the Birbaumer group trains on brief transients (nominally 8 seconds), whereas we train on the continuous signal. We have simply gone from a small duty cycle to a large one. Giving the brain more signal to engage with turns out to engage the brain more actively. This should be no surprise. At a minimum, thorough-going skepticism regarding our findings seems bizarre given the long prior history here of effective training on slow cortical potentials in Germany.

If on the other hand the work is taken seriously then a number of interesting questions open up. It has been observed on many occasions that the energy consumption of the brain is only marginally increased when we are actually engaged in purposeful mental or physical activity. In conventional functional imagery one typically has to do a careful comparison of the brain under task with the brain in baseline in order to extract the small signal of interest. The brain’s activity level also remains quite high even during our sleep. The low-frequency signal we are training likewise appears to be present whether we are active, merely awake, or even asleep.

It could be that at the lowest frequencies we are training brain rhythms that are foundational for the higher EEG frequencies that organize our activities more specifically. Our EEG may to an extent be hierarchically constructed in the frequency domain, with the nesting of higher-frequency activity within the lower. It is not necessary for that to be the case in order to explain our work. Even if the low-frequency activity were entirely an epiphenomenon of the higher-frequency realm the neurofeedback could still “work.” Finally, even if the signal we are tracking has a primary non-neuronal origin, the neurofeedback can still be an effective clinical strategy.

One thing we can argue from fundamentals is that the observed fluctuation in the electro-chemical potential transiently modulates the excitability of neuronal populations. So even if the signal of interest is not entirely traceable to neuronal origin in first instance, it nevertheless is intimately involved with our basic task, which is to bring the activation-relaxation dynamics of neuronal networks under improved self-regulatory control. And in this regard we find ourselves very much where EEG neurofeedback has always been.

Most of neurofeedback has historically been concerned with the resting states of the cortical system–the alpha and the SMR rhythm, and to a lesser degree frontal midline theta. Training to enhance SMR and alpha amplitudes promotes disengagement and lowered levels of excitability. We could consider them as part of the brain’s “preparedness state,” its readiness to function. The lowest EEG frequencies we are working with now could be an essential part of that maintenance of readiness to function. A good deal of research has recently gone into describing the “default network” that organizes the EEG in baseline. The lowest-frequency EEGs could provide the essential large-scale order for such a global default network. By the same token, they could help to provide the continuity necessary for the concatenation of brain states that makes up our experience of the world. We may indeed be finding ourselves at the threshold of a very exciting new frontier for neurofeedback.

The interested reader is invited to see the following article on infra-low frequency training:

Siegfried Othmer, Ph.D.

Leave a Reply