Biofeedback of Brain Wave Activity

J.P. Rosenfeld,Ph.D.

Department of Psychology

Northwestern University

Evanston, IL, USA 60208

phone:847-491-3629

fax: 847-491-7859

email: jp-rosenfeld@nwu.edu

 Acknowledgment. I am grateful to a succession of talented Collaborators over the years, especially Karl Diedrichs, Bruce Hetzler, Paul Birkel, and Bob Dowman. I am also grateful to Steve Fox and Al Rudell for Early inspiration.

In some ways, brain wave feedback (BWB) is like other

biofeedback, but in many other ways, BWB is unique. In general, biofeedback

methodology presents to a subject's conscious perceptual sensorium a representation of the activity of a given organ-system, of with the subject is ordinarily unaware. For example, one usually has no innate awareness of his moment-to-moment blood pressure, but it can be recorded and displayed for conscious perception. If it is pathologically high, the subject needs to be able to know its value on a moment-to-moment basis in order to become aware of which of his cognitions and/or behaviors cause it to move in a healthy direction. Some BWB is perfectly analogous. Epileptic seizure activity and seizure-related activity, for example, is directly represented in scalp-recorded brain waves (EEG). It may represent pathological brain activity, exactly as recorded blood pressure represents pathological cardiovascular activity. Moreover, as Sterman, Lubar, Finley, and others have shown (in other chapters of this volume), BWB can be used to represent pathological EEG to a subject, who is (ordinarily) unaware of his EEG frequency spectrum, so that he may learn to generate those cognitions/emotions/behaviors whose effects tend to normalize

the EEG. In this example in which BWB is like other biofeedback, the

fed-back EEG information directly represents the action of the diseased

organ-system, i.e., the central nervous system.

However the fact that, more generally, the central nervous system (CNS)controls and/or represents the activities of most(if not all) peripheral systems, voluntary (biofeedback) control of CNS activity has futuristic-sounding but logically real implications for control of many other systems in a more direct manner than that afforded by more familiar, direct, peripheral biofeedback methods. Equally important and fascinating is the fact that since psychological functions -- cognition, emotion, perception -- are probably uniquely mediated by brain systems, BWB must afford the most direct and perhaps the only method for drug-free therapeutic management of psychological disorders, or for tuning and improving normal psychology.

2. Trivial Mediation

There is another side of this potentially valuable coin, however: Since

the CNS is probably connected to all other behavioral-peripheral systems,

in order to make the case that there is the potential for unique effects of

BWB, one has to demonstrate that various instances of BWB do not involve mere central representations of known peripheral phenomena. This is known as the "trivial mediation" issue, and an example of it may be found by briefly re-considering our demonstrations years ago that ensory-evoked EEG potentials can be voluntarily controlled (see, e.g., Rosenfeld & Hetzler, 1979). We trained cats, people, and rats to both increase and decrease the size of components of cortical EEG potentials evoked by stroboscopic light flashes or electric shocks presented about every 3-5 seconds. The subjects were readily able to alter the evoked potentials (EPs), but how did they do so? More precisely we needed to rule out possible trivial things that they could have done which would result in altered EPs. For example, since the size of flash EPs is a monotonic though noisy function of light intensity impinging on subjects’ eyes, subjects could increase flash Eps by orienting to the light source, or decrease EPs by facing away, or by closing eyes. Also, it is known (from some of our own work, e.g., Rosenfeld & Fox, 1972) that voluntary movement generates somatic sensory neuronal activity which ascends the CNS and affects cortical activity in widespread cortical and subcortical areas. Thus, our test animals could alter flash-EPs by altering receptor orientation and/or by moving in such a way as to affect EP amplitude size. We have previously detailed a 10 year long series of experiments designed to rule out such trivial mediation (reviewed by Rosenfeld et al. 1985; Rosenfeld 1974, 1977), and it is not appropriate to review all these experiments here. I will describe just one study (Rosenfeld, Hetzler, Birkel, Antoinetti, & Kowatch, 1976) which went a long way in terms of ruling out trivial mediation of visual EP-BWB in visual cortex.

First, the visual EPs (in rats) were evoked by non-painful electrical stimuli to the visual pathways via electrodes implanted in optic tracts. The rats were run in total darkness. The EP component specified for alteration appeared .07 seconds (70 milliseconds) after the electrical stimulation applications, which were delivered at intervals randomly varying between 3 and 5 seconds. Movement of subjects was monitored at all times by a highly sensitive transducer mounted directly on the rats. Beginning 300 milliseconds after a visual stimulus for a successful trial, the rats were rewarded with a brain stimulation in the lateral hypothalamic "pleasure center.

Clearly, the highly successful rats could not have been altering their orientations to the stimuli, which were delivered at electrodes fixed in their brains to activate the same optic tract location on each trial, every day. Altering orientation to background light was prevented by the running of rats in total darkness. Discrete movements timelocked to the interstimulus intervals were impossible because these intervals randomly varied. The timuli themselves could not serve as timing cues for timelocked movements because the latency of the EP component specified for change was (at 70 milliseconds) significantly less than behavioral reaction time to the stimuli. It was possible to rule out neural effects of non-timelocked movements as trivial mediators of the BWB in two ways: First, the changes observed in the EPs were highly specific, i.e., usually localized to the one late component of the EP prespecified for training, whereas ongoing body movements are known to affect all components of the visual evoked potential. Second, a detailed analysis of the recorded ongoing movements showed no differences between the movement patterns associated with BWB-trained increases in EPs and the movements associated with trained decreases in the EPs (see also Hetzler, Rosenfeld & Birkel, 1978).

The inescapable conclusion is that the rats were probably learning to produce specific mental activities—the rat versions of thoughts or feelings—which altered the cortical synaptic organizations where the specified EP components were generated. Moreover, there were no outward signs (behaviors) of the particular cognitive activities. It should be noted that Dowman & Rosenfeld (1985), later demonstrated the same effects using a completely different sensory system (the trigeminal tract) as a stimulation site, an EP recorded from somatosensory as opposed to visual cortex, and a multilevel behavioral analysis.

Voluntary control of brain activity with no immediate outward signs may sound far fetched, however, it can be easily rationalized and made intuitively obvious by pointing out that such is what happens all the time as humans sit, quietly thinking, reacting internally to stimuli, processing information, experiencing emotions, and so forth. Indeed, in another line of research, we have shown that certain "endogenous" evoked EEG waves (or "Event-Related Potentials, ERPs) can be used as mind-reading lie detectors in subjects motivated to show no outward signs of their recognition of key stimuli (Rosenfeld, Angell, Johnson, & Quian, 1991; Rosenfeld, Sweet & Ellwanger, 1994). Indeed, we have come to believe that BWB in our hands involves production of neural responses well within the normal repertoire of neuronal activity (although see Rosenfeld & Hetzler, 1973). We believe this because the criteria we define for successful trials (hits) are based on the normal distribution of neuronal responses prior to training. E.g., we collect the pre-training amplitude distribution and define a hit as an amplitude value greater than .5 to 1 standard deviation(s) from the pre-training mean. Thus a successful subject is simply generating more instances of a normal event which had a lower probability of occurrence prior to training.

Is it always necessary to go to great lengths to prove that a given instance of BWB is free of trivial mediation? The answer to this question depends on the aims of the BWB. Clearly if one wants to make a claim that there is something unique and novel about a BWB effect, it would seem incumbent upon one to show that the BWB was unmediated by something previously known and familiar. For example, we have argued (Rosenfeld & Hetzler, 1979) that EP biofeedback may provide a novel (motor-free) response system with which to test the generality of putative learning laws. It becomes obviously essential, given this aim, to demonstrate clearly that the BWB is indeed unmediated by the familiar voluntary motor behavior traditionally utilized in conditioning research. On the other hand, if one’s aim is mostly clinical-therapeutic, one is less interested in mediation mechanisms than in positive clinical results. Sometimes, however, it would seem that a strong, positive clinical effect

of BWB on a pathological condition previously refractory to all kinds of

alternative interventions, is itself the best argument for the novelty of

the BWB utilized.

3. Delayed Effects of BWB: Clinical Effects, BWB and Pain

It was stated above that some instances of non-trivially mediated BWB would involve CNS changes with no immediate outward signs. On the other hand, a positive clinical effect by definition must involve an outwardly expressed, measurable effect at some point. This could simply be a verbalized expression of inner experiences—"I feel less pain" or "I am less depressed." Or it could involve a change in measurable behavior such as spastic activity, convulsions, gait, etc. It is well known that simply causing a change in activity at a given CNS locus can affect the manner in which synaptic organizations at that locus transmit neural information in the future. Experimental models of this effect include Long Term Potentiation (LTP) and Post-Tetanic Potentiation, (Bliss & Lynch, 1988; Colley & Routtenberg, 1993). In LTP, for example, a nerve fiber bundle afferent to rat hippocampus is electrically stimulated repeatedly. The synapses involved are subsequently found to have a lower activation threshold than prior to stimulation. This effect, which has been shown to have membrane structural-biochemical correlates, has been suggested as a simple model of learning, defined as a change in behavior produced by experience (Bliss, & Lynch, 1988; Colley, & Routtenberg, 1993). It is not a bad model for clinical effects of BWB, either. Of course, in the case of BWB, it is the learned alteration in EEG, rather than brain stimulation which is hypothesized to effect the synaptic change.

For example, we performed several studies in rats and in humans, in which we trained subjects to change EPs evoked by non-aversive electric pulses applied to the somatosensory system; the trigeminal tract in rats or the skin surface (back of hand or forearm) in humans. The human subjects confirmed verbally that the stimuli felt like just noticeable (and certainly not painful) skin taps. We inferred the painlessness of the stimuli in rats from the fact that they would not stop eating during stimulation. Indeed, they mostly showed no outward signs at all of the stimulative experience, and our only way of knowing for certain that the stimuli were activating the CNS was via observation of the stimulus-evoked cortical potentials. In rats, the EPs were monopolarly recorded from electrodes placed directly over the primary somatic sensory cortex; the reference and grounds were in the frontal sinus. In humans, EPs were recorded from the vertex (Cz), a site over human sensorimotor cortex, referenced to linked mastoids, with the forehead grounded. We expected and confirmed (Rosenfeld et al., 1985; Dowman & Rosenfeld, 1985a, 1985b; Rosenfeld, Dowman, Heinricher, & Silva, 1984) that the effects of this BWB of somatic-sensory evoked cortical potentials would involve changes in subjects’ pain sensitivities (measured after BWB), even though the EPs were not evoked themselves by painful stimuli. Our expectation was based on the fact that the same somatic-sensory cortical neurons subserving non-noxious somatic sensory submodalities (e.g., touch) also respond to painful stimuli. (These well studied cells are commonly referred to as "WDR" cells wide dynamic range neurons; see Rosenfeld et al., 1984.)

The changes in pain sensitivity we actually did observe were quite sizeable; equivalent to the effects of moderate doses of morphine. They were probably also somatotopically specific—only the side of the face represented in the conditioned side of the cortex showed effects. They were also related to endogenous opioids (Enkephalin, Endorphin, etc.), since naloxone antagonized the BWB effects on pain perception. We reasoned that all these dramatic observations occurred because by conditioning WDR neurons’ responses to innocuous stimuli, we effected an alteration in their responsiveness to noxious (i.e., painful) stimuli with which they were challenged following the 33 minute conditioning blocks of 500 trials each.

Incidentally, it should be appreciated that although the Eps are evoked by electrical stimuli of constant intensity, the BWB-produced changes in the EPs are effected endogenously (by the subject). The protocol in humans was as follows: An evoking stimulus (see above) was presented every 4 seconds. The Cz-recorded EPs were led to a PDP8 minicomputer via a Digital Equipment Corporation A/D converter, after amplification with a Grass P5J amplifier set to pass signals between 3 and 300 Hz. The A/D sampled at 1 KHz. On a baseline day, in which 250 trials (stimulus presentations) occurred, the average monopolar EP at Cz referenced to mastoids, was computed. In particular, the average of a 100 msec segment (criterion segment) of EP occurring in a positive component peaking at about 200 msec post-stimulus was determined on each trial, and these segment averages were then averaged over all trials. The standard deviation (SD) of the criterion segment was also computed. In subsequent training days, a reward tone was presented immediately following hit trials, defined as those on which the criterion segment was .7 SDs greater than the pre-training mean. (This protocol has since been implemented on an MS-DOS-based PC clone; interested readers are welcome to contact me for further information.)

BWB Alpha Asymmetry Effects of Emotion and Depression

Our working model for very recent work in BWB with possible applications for depression is related, but not similar, to the model we used in our just described pain-related BWB. The approach to be now described has as its conceptual foundation the work of Davidson and colleagues (Davidson, 1992). This research has elegantly extended earlier neuropsychological data suggesting that there are emotion-relevant systems in the frontal cortical areas of brain. In particular the right frontal cortex seems to contain a system which mediates avoidance reactions associated with negative emotional experience and, in contrast, the left frontal cortex appears to contain a system specializing in elaborating approach reactions facilitating positive experience. Thus relative increased right frontal activation, which can be indexed by reduced right alpha activity, would be expected to go with negative reactions, whereas enhanced left frontal activation, expressed in reduced left alpha power would go with positive experience. (Note activation is inversely related to alpha activity.) Although it is possible to distinctly record left and right alpha, for technical reasons beyond the scope of this chapter, Davidson and colleagues have utilized a right-minus-left log alpha power index (called "ZA" here) as an indicator of emotion. Thus if ZA is high, there is relatively greater left frontal (approach) activation than right frontal (avoid) activation, and the corresponding emotional state would be positive. By parallel reasoning, low ZA would be expected to go with negative emotional states. The Davidson group has confirmed these expectations in many ways (see Davidson, 1992, for a review), but most relevant to present concerns, they have shown that currently and previously depressed individuals have resting ZA scores on the low end of the ZA population distribution, whereas non-depressed individuals have moderate to higher levels of ZA (Henriques & Davidson, 1990). Davidson (1992) has argued that although some of his other work indicated that ZA can index temporary states induced by experimental emotional manipulations, the fact that previously depressed (but then successfully treated and currently non-depressed) persons still show the characteristically lower ZA scores indicaates that ZA may be a stable marker of the lifelong trait of vulnerability to depression.

There are obvious clinical implications here of subjecting ZA to BWB. Our view is that such BWB might be a beneficial and non-intrusive intervention for depressive disorders, with none of the unpleasant side effects which can accompany traditional treatments for depression such as ECT and pharmacotherapies. I suspect though that future research will show that the best time to intervene with BWB of ZA will not be in the depths of depression, but when known vulnerable individuals are in stressful periods which often bring on depressive episodes. (This notion is based on the often-expressed view that severe depression is a counterindication for any biofeedback; (Schwartz, 1987, p.170). BWB of ZA might even be a useful regular prophylactic activity. This, of course, is looking far ahead into the future. What is the status of this project now?

We have done three studies of five normal subjects each, mostly to learn 1) whether or not such ZA training is possible, 2) how readily achievable (or how difficult) it is, and 3) what some of the training parameters should be (Rosenfeld, 1994, 1995, Rosenfeld et al., 1993). Our first study was done by using active filters to extract and index alpha power. Our more recent work has utilized Fast Fourier Transforms to do the alpha powerextraction and BWB (with our own software running on a PC386 clone running at 40 MHZ and equipped with a mathematics co-processor). We have found that half to two-thirds of our normal subjects are able to learn to increase ZA. Our mastery criterion for learning was a doubling of mean hit-rate from that achieved during the first training day. I hasten to note here that, due to limited resources, we could utilize only three (one hour) training sessions following a baseline and exploration session in each subject. Since other clinical BWB paradigms involve typically more than ten times this number of training sessions, we would expect a higher percentage of trainable persons in the future when we too can utilize multiple sessions. Our sessions consists of 400 "trials," i.e., FFT epochs of 1 second of analysis. Also, our method of reinforcement so far has been quite primitive (sounding high tones of one fixed frequency when the trial has been a failure, sounding low tones of another fixed frequency for hits, and no sounds during artifact trials). The protocol will probably work much better when we give moment-to-moment feedback such that either volume or pitch continuously changes in proportion to the instantaneous asymmetry score, with "hit trials" additionally signaled with a chime. We have recently reported success with one clinical patient using continuously variable tone volume and 26 training sessions (Rosenfeld, Baehr, & Baehr, 1995). We also need to further explore appropriate epoch (trial) durations, intertrial intervals, and so on, before the potential of the paradigm can be fully realized. Nevertheless, our success

rate, given our preliminary conditions, is encouraging.

Two other results should also be noted: 1) Although until recently, we have not systematically explored behavioral-affective changes, associated with BWB of ZA, recent informal monitoring of two recent subjects on closed circuit TV during training revealed profound emotional changes confirmed by post-training debriefing. The changes appear paradoxically puzzling—grief and tears accompanied increased ZA production which, in Davidson’s hands, indexes positive emotional effects. It is tempting to speculate from these data about possible rebound effects as complex emotional phenomena, but at this stage of the research, I feel it is more appropriate to devote energy to more empirical work. 2) Changing ZA, logically, can result from three effects: increased right alpha power, reduced left alpha power, simultaneously increased right and decreased left power. We have seen all three of these effects in various subjects. Most important: some subjects start with a "depressed" ZA score, i.e., more left than right alpha power, but after training, show more right than left alpha power. In other words, in the course of training, the left and right power curves cross one another going in opposite directions. Such subjects both reduce left alpha and enhance right alpha power at the same time. This bodes well for working with depressed patients in the future, for it suggests that people can go from a depressed to a normal asymmetry profile via BWB.

Another implication of our preliminary results (that it is possible to voluntarily control ZA) is the possibility of testing to what extent ZA is a trait marker of vulnerability to depression. If long term changes in ZA, resulting from BWB, can be demonstrated, and if these correlate with emotional changes, then ZA and/or vulnerability to depression may not be stable traits, which is good. On the other hand, if further research reveals that non-depressed but low baseline ZA subjects cannot modify ZA as readily as baseline-high ZA subjects, the trait hypothesis will receive strong support.

Our model of the ZA-BWB is of necessity different than that described earlier regarding our pain studies. This is because whereas the neural events altered in the pain studies involved evoked activities, the neural events from which ZA is extracted are ongoing, spontaneous EEG waves. Nevertheless it is still possible that increased ZA events occurring during BWB facilitate such events happening in the future via some potentiation process (like LTP). This view will become problematic, however, if we continue to see negative emotional concomitants of increased ZA in training, followed by later rebounds in the direction of positive affect. Such results might suggest a receptor regulation-based process (as happens in pharmacotherapy with depressed persons) involving a slow build-up or rebound of neurotransmitter receptors and/or substances related to synaptic transmission. These notions approach the borderline of wild speculation, but hopefully have some heuristic value in suggesting very much needed future research. Finally, it may be noted that most biofeedback protocols are not pure when they reach clinical maturity. They may then include ancillary or supplementary procedures such as cognitive and behavioral training, relaxation procedures, and so on. It is not unexpected that such will also be the case for the BWB protocols described here.

References

Bliss, T.V.P., & Lynch, M.A. (1988). Long-term potentiation of synaptic transmission in the hippocampus properties and mechanism. In Landfield, P.W., & Deadwyler, A. (Eds.), Long-Term Potentiation: From Biophysics to Behavior, Alan R. Liss, New York, 3-72.

Colley, P.A., & Routtenberg, A. (1993). Long-term potentiation as synaptic dialogue. Brain Research Reviews, 18, 115-122.

Davidson, R.J. (1992). Anterior cerebral asymmetry and the value of emotion. Brain and Cognition, 20, 125-151.

Dowman, R. and Rosenfeld, J.P. (1985a). Operant conditioning of somatosensory evoked potential (SEP) in rats. I. Specific changes in SEP amplitude and a naloxone-reversible, somatotopically sepcific change in facial nociception. Brain Research, 333, 201-212.

Dowman, R. and Rosenfeld, J.P. (1985b). Operant conditioning of somatosensory evoked potential (SEP) in rats. II. Associated changes in reflex and continuous non-timelocked movements. Brain Research, 333, 213-222.

Henriques, J.B. & Davidson, R.J. (1990). Regional brain electrical asymmetries discriminate between previously depressed and healthy control subject. Journal of Abnormal Psychology, 99, 22-31.

Hetzler, B.E., Rosenfeld, J.P., and Birkel, P.A. (1976). Operant control of central evoked potentials in animals, with detailed analysis of ongoing body movements. Psychophysiology, 14, 91.

Hetzler, B.E., Rosenfeld, J.P., Birkel, P.A., and Antoinetti, D.N. (1977). Characteristics of operant control of centrally evoked potentials in rats. Physiology and Behavior, 19, 527-534.

Hetzler, B.E., Rosenfeld, J.P., and Birkel, P.A. (1978). Analysis of body movements during operant control of central evoked potentials. Physiology and Behavior, 21, 1047-1050.

Rosenfeld, J.P. and Fox, S.S. (1972). Sequential representation of voluntary movement in cortical macropotential: Direct control of behavior by operant conditioning of wave amplitude. Journal of Neurophysiology, 35, 879-891.

Rosenfeld, J.P. and Hetzler, B.E. (1973). Operant controlled evoked responses: Discrimination of conditioned and normally occurring components. Science, 181, 767-770.

Rosenfeld, J.P. (1974). Evoked potential conditioning in neuroscience research. In: M. Chase (Eds.), Operant Control of Brain Activity, (Perspectives in the Brain Sciences, Vol. II), Brain Information Service, Brain Research Institute, UCLA.

Rosenfeld, J.P. and Rudell, A.P. (1976). Mediation of operant controlled neural activity. In: D. Mustovsky (Ed.), Behavior Control and Modification of Physiological Activity, New York: Appleton-Century-Crofts.

Rosenfeld, J.P., Hetzler, B.E., Birkel, P., Antoinetti, D., and Kowatch, R. (1976). Operant conditioned potentials, centrally evoked at random intervals. Behavioral Biology, 16, 305-317.

Rosenfeld, J.P. Conditioning changes in the evoked response. (1977). In: G.E. Schwartz and J. Beatty (Eds.), Biofeedback: Theory and Research. New York: Academic Press, 377-388.

Rosenfeld, J.P. and Hetzler, B.E. (1979). Significance and mediation of neural and other biofeedback. International Journal of Neuroscience, 8, 233-250.

Rosenfeld, J.P., Dowman, R., Heinricher, M., and Silvia, R. (1984). Operantly controlled somatosensory evoked potentials: Specific effects on pain processes. In: Self-Regulation of the Brain and Behavior, Ed. by B. Rockstroh, T. Elbert, W. Lutzenberger and N. Birbaumer, Berlin: Springer-Verlag, 164-179.

Rosenfeld, J.P., Stamm, J., Roger, M., Birbaumer, N., Rockstroh, B., and Elbert, T. (1983). Biofeedback of event-related potentials. In: Karrer, R., Cohen, J., and Tueting, P. (Eds.), Brain and Information: Event-Related Potentials, Proc. VI Int. Conf. on Event-Related Slow Potentials of the Brain. N.Y. Acad. Sci., Monograph #12, 653-666.

Rosenfeld, J.P., Silvia, R., Weitkunat, R., and Dowman, R. (1985). Operant Control of Human Somatosensory Evoked Response Alters Experimental Pain Perception. Advances in Neurology, 9, Raven Press, 343-348.

Rosenfeld, J.P., Angell, A., Johnson, M., and Qian, J. (1991). An ERP-based control-question lie detector analog: Algorithms for discriminating effects within individual waveforms. Psychophysiology, 28, 320-336.

Rosenfeld, J.P., Cha, G., & Blair, T. (1993). Operant (Biofeedback) control of left-right frontal EEG alpha power differences. Psychophysiology, 30 (Supplement 1), p. S. 56.

Rosenfeld, P., Sweet, J., & Ellwanger, J.P. (1993). P3 and concealed information in the identification of simulated amnesia. Psychophysiology, 30 (Supplement 1), p. S. 56.

Rosenfeld, J.P. (1994). Method and system for treatment of depression with biofeedback using left-right brain wave asymmetry. U.S. Patent 5,280,793 issued 1-25-94:

Rosenfeld, J.P. (1995). Method and system for modification of condition with neural biofeedback using left-right brain wave biofeedback. U.S. Patent 5,450,855 issued Sept. 19, 1995.

Rosenfeld, J.P., Baehr, E., and Baehr, R. (1995). EEG frontal alpha asymmetry measures correlate with mood change scores during asymmetry biofeedback. Proc. 26 Ann. Meeting, Assoc. Applied Psychophysiol. Biofeedback, Cincinnati, Ohio, pp. 125-128.

Schwartz, M.S. (1987). Biofeedback: A practitioner’s guide, N.Y., Guilford Press, p. 170.

 

Florence S. Sales

Graduate Admissions Coordinator

Northwestern University

Department of Psychology

Evanston, IL 60208

 

J.P. Rosenfeld,Ph.D.