Neurofeedback: Exercise or Operant Conditioning?

I believe neurofeedback is like exercise. In most exercise we have some form of mirror: you might be watching yourself trying to achieve and stay in a pose in yoga, or have a pulse-meter that informs you when you are in or out of the training range in aerobics. There’s nothing innately reinforcing about the pulse meter or even looking at yourself in a mirror or noticing how many pounds you are lifting and how many times. They are ways of helping the brain recognize that it is moving in a desired direction. I suppose you could argue that a pulse meter beeping when you are above or below the aerobic training range is conditioning a response, but maybe it’s just telling you to speed up or slow down when you are walking on a treadmill.

Per the literature asserting that neurofeedback is a conditioned response, positive feedback is much more powerful than negative, and continuous is more powerful than contingent.  However, I often use “negative” feedback when training a client during task–a tone that plays like an alarm when the client’s theta (for example) rises above target and silence when it is within range.   It has worked very well.  And I’m a strong believer that the type of feedback (continuous or contingent) should be related to the type of exercise we are asking the brain to do. Activating (aerobic) seems to work best with contingent (on/off) feedback.  Continuous feedback works better for brain yoga/stretching (de-activating).   I have no idea, if NF is conditioned response behaviorism, why this would be true, but it makes a lot of sense if training is brain exercise.  Even the issue of how long to train in a segment fits with exercise. If you are trying to build stamina with aerobics or weight training, you use shorter sessions of pushing your limits separated by breaks (like wind-sprints). If you are trying to release something that is too tight, you use longer, slower training segments and let the unwinding happen little by little.

The nice thing about this approach is that you don’t have to worry about whether the feedback is intrinsically rewarding or not (so you don’t need different feedback for each client). It is, as someone suggested in a previous post, and I have long said) simply a mirror.

Classical and Operant conditioning are two concepts that behavioral psychologists use to explain what happens from their point of view. Their point of view is far from the only one–or even the best one–but since most literature about neurofeedback has been written by psychologists–and most of them are trained behaviorally–pretty much any article you read about neurofeedback will start off by taking as a given that training is “operant” conditioning. NOT classical. It’s a handy shorthand that underlies the concept that neurological is a psychological intervention rather than something we do for ourselves. But it is far from being proven.

First, what is classical conditioning? Pavlov’s dogs who salivated at the sound of a bell. It involves placing a “neutral” stimulus BEFORE a REFLEX. A hundred years ago John Watson used a hammer to bang a metal pipe, thus producing a fear response in a baby boy who was not previously afraid of small furry animals. After doing this often enough–hey, this was Science, so it was okay–he managed to get the little child to cry in fear whenever any furry animal was presented to him. That’s certainly worth doing. Save a lot of money on pet food over the years. CC is reportedly useful in desensitizing or flooding to extinguish a fear response, though after a century it’s still not a primary approach. It’s supposed to be good in smoking cessation or intervening in alcohol abuse, though again, I’m not sure how often it’s used. Instead of making you vomit when you take a drink, brain training creates enough changes (painlessly) that the compulsion to drink is no longer there–not simply scared away.

In any case CC is pretty far from what we do in brain training, where we provide FEEDBACK–not a neutral stimulus–AFTER the brain does something. Unless you worked pretty hard to play a certain note over and over just before the brain produced a burst of SMR, for example, it’s unlikely there would be much of an effect.

Operant conditioning (OC) is a bit different. It involves rewarding or punishing AFTER a VOLUNTARY behavior to reinforce or extinguish it. Teach a pigeon to peck a button by giving it food every time it pecks a red light–not a green light. Or Barry Sterman provides sweetened milk drops to cats whenever they produce a burst of SMR above a certain level.

If you are a hungry pigeon or cat, some bird seed or sweet milk is pretty obviously a reward, whereas, for example, some beeps from a computer, or momentary darkening of a screen may not be so clear. It raises the questions such as: does the client “LIKE” the feedback? It would seem that disliked training should have REDUCED the level of the desired behavior, instead of increasing it!

Another problem–one raised by Sterman himself–has to do with the discreteness of the event and the feedback. Since bursts of, say, SMR are being produced multiple times per second–among many other types of activity–how exactly does the brain know which event is resulting in the reward? Sterman made quite a point a number of years ago of telling trainers they needed to give a recharge period after each reward. Feedback had to be intermittent rather than continuous. After a burst of SMR over threshold and production of a reward, you had to stop giving feedback for a time and then start again looking for the SMR burst. It was pretty cumbersome way of training, though (since it was Sterman telling us to do it) more than a few people went to the trouble of trying to produce designs that more or less met the requirements. To the best of my knowledge people working with clients didn’t notice any particular improvement in their clients’ ability to produce SMR–much less their ability to relax physically, so it’s pretty rare to find anyone still trying to do this.

Maybe behavioral psychology’s attempt to cast neurofeedback in its own image requires some pretty major leaps of faith to make any sense. But gosh, if brain training ISN’T OC, then what could it possibly be?

What if brain training isn’t psychology? After all, it works with spiritual and physical as well as mental and emotional issues. What if it isn’t about fixing mental disorders but about moving toward greater flexibility and range. Is meditation psychology? Are brain exercises like Lumosity psychology? What if brain training isn’t about “conditioning” the brain in some kind of mechanistic way but instead is about giving it greater control instead of less?

The hint is in the name: Feedback. One way we learn to control ourselves is by using mirrors. We do something and the feedback shows us HOW well. It seems kind of simple, but the reality is that the human brain learns most everything from mirrors. It responds to sensory inputs by taking action, then it sees what happens. Feedback. That might help explain Birgit’s client’s positive response to unpleasant feedback. Even if I don’t LIKE what I see in the mirror, I can still learn from it. And taking the mirror home and looking into it when you aren’t trying to do anything isn’t likely to be terribly productive.

I don’t care whether you think the behavioral psychology model is more useful or the feedback/mirrors model, or just an “exercise” model. It means almost nothing if your main focus is on guiding a client to produce changes. I just hate to have us tossing around concepts like “conditioning”–which suggest that something is being done TO us–without really thinking about them. Exercising in front of a mirror is something we do FOR ourselves. I like that approach much better.

Moreover, systems as complex as the nervous system are almost certainly organized around principles of oscillatory “ordered chaos,” rather than being neat, linear systems (linear in both a mathematical and a metaphorical sense).  Thus, neurofeedback may be as much about resetting equilibrium points in the dynamic chaos of functional neurological systems as it is about teaching certain groups of neurons (visualize students sitting in neat rows in a classroom) to behave better.  Obviously, it’s easier for us to think in linear terms (e.g. linear algebra versus partial differential equations), but that doesn’t mean that the brains that do that thinking are linear oscillatory systems.  All that I am suggesting is that we might be doing something similar with neurofeedback:  not so much “teaching” the brain to “behave” better as cajoling it into different states of chaotic equilibrium.