ALPHA WAVE ENHANCEMENT AND THETA WAVE SUPPRESSION IN THE
CONTROL OF EPILEPTIC SEIZURES
GEORGE EDWARD GERCKEN
A DISSERTATION PRESENTED TO THE GRADUATE COUNCIL OF
THE UNIVERSITY OF FLORIDA
IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE
DEGREE OF DOCTOR OF PHILOSOPHY
UNIVERSITY OF FLORIDA
In the romantic imagination of the public, scientific progress
is pictured as a single individual stumbling into some discovery
while toiling late and alone in a makeshift laboratory. In reality,
scientific progress is the sum total of the tiny and painstaking
contributions of generations of researchers. It seems fitting, there-
fore, that before beginning this report of my own small contribution
to the therapeutics of epilepsy, I should acknowledge the encouragement,
assistance, and cooperation of the many colleagues without whom this
work would not have been possible.
I am particularly grateful to the members of my committee:
Drs. Hugh C. Davis, Cynthia Belar, Robert Isaacson, Warren J. Rice,
and B. Joe Wilder. Dr. Davis, the chairman, stepped in unselfishly
to fill the void left by the sudden death of Sidney Jourard. Dr. Belar
loaned us the seemingly unloanable 5400 computer at a time when ours
had given up the ghost. Dr. Wilder's talk at the Michigan Epilepsy
Center in 1973 convinced me to come to the University of Florida.
Dr. Isaacson kindly shared his profound professional expertise and his
unerring eye for precision and detail. A special thanks to Dr. Warren
Rice, my cochairman, who contributed time, effort, and understanding
for which I can never fully express my gratitude.
I should also like to express my deep appreciation and gratitude
to the Veterans Administration for its excellent support, and
especially to the staff at the Jacksonville outpatient clinic.
My sincerest thanks to Dr. Jack Smith and Jose Principe who
provided valuable computer access and programming efforts in the
analysis of my data.
Finally, to Dr. L. Jim Willmore, whose function as friend,
counselor, and colleague helped me flush out the real from the
fantastic in moments when I wasn't sure, a special thanks.
This project has been simultaneously frustrating, fascinating,
exhausting and stimulating. I hope as well that it will serve as a
stepping stone for other researchers exploring the new and intriguing
field of physiological feedback.
TABLE OF CONTENTS
ACKNOWLEDGEMENTS . . .
ABSTRACT . . .
I A REVIEW OF THE LITERATURE . .
Definition of Epilepsy . .
Medical Treatment of Epilepsy .
Learning-Theoretic Models of Epilepsy .
Learning Theory Models . .
Inhibitory Synchronous Activity and Seizure
Physiological Feedback .
The Initial Use of Physiological
Treatment of Epilepsy .
Recent Experimentation .
Summary and Hypotheses .
II METHOD . .
Subject Selection .
Individual Subject Profiles. .
Procedure . .
Experimental Design .
Apparatus . .
Data Analysis .
III RESULTS . . .
Data with Respect to Hypothesis 1 .
Data with Respect to Hypothesis 2 .
Data with Respect to Hypotheses 3 and 4 .
Data with Respect to Hypothesis 4 .
Between Slope and Mean Differences .
IV DISCUSSION . . .
Discrepancies Between Hypotheses and Results .
Design Critique and Suggestions for Improvement.
Suggestions for Future Research .
A TRAINING TAPE SCRIPT . .
B CHI SQUARE PLOT OF OBSERVED VS. EXPECTED SEIZURE
FREQUENCIES BY CONDITION . .
C SUMMARY TABLE OF SIGNIFICANT SLOPES. .
D SELECTED PAIR-WISE CONTRASTS BETWEEN BASELINE AND
TREATMENT CONDITIONS FOR SLOPES AND LEVELS .
REFERENCES . . .
BIOGRAPHICAL SKETCH . .
Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy
ALPHA WAVE ENHANCEMENT AND THETA WAVE SUPPRESSION IN THE
CONTROL OF EPILEPTIC SEIZURES
George Edward Gercken
Chairman: Hugh C. Davis
Major Department: Psychology
The present research was concerned with investigating the EEG
physiological feedback technique of combined alpha wave enhancement,
theta wave suppression, and its effect upon the attenuation of epilep-
tic seizures. A review of the literature indicated a possible link
between midrange synchronous EEG activity, chiefly in the alpha range,
and inhibition of seizures. In addition, the review demonstrated a
possible connection between slow wave hypersynchony, chiefly in the
theta range, and seizure propagation in focal seizure patients.
Five subjects, four male and one female, were employed in the
study. All subjects were classified as having uncontrolled focal
seizures of at least five years' duration. Each subject was required
to maintain a seizure diary one month prior to and throughout the
course of the experiment. All subjects were administered EEGs using
the training referents T3-T4, C3-C4 before training was begun and
immediately after completion.
Following an A-B1-C-B2-(D) research design (A=baseline, Bl
contingent feedback 1, C=noncontingent feedback, B2=contingent
feedback 2, D=theta suppression), subjects were trained to enhance EEG
alpha wave and suppress theta wave production.
The results indicated support for the hypothesis that enhancement
of synchronous midfrequency activity had a significant effect upon
seizure rate. Discrepancies between these results and those of
Sterman were discussed in terms of homogeneity of seizure type. The
greater variability of seizure classification of Sterman's subjects
made his results difficult to interpret from an etiological standpoint.
The present study's concentration only on focal seizure patients
increases the specificity of results for this subgroup.
Methodological problems were discussed in terms of extending
baseline data or interjecting a multiple baseline approach to more
accurately gauge the effectiveness of training procedures. The
extension of the noncontingent condition was recommended to permit a
more direct assessment of treatment effects.
A REVIEW OF THE LITERATURE
The sky spun like a mighty wheel;
I saw the trees like drunkards reel
And a light flash sprang o'er my eyes,
Which saw no farther. He who dies
Can die no more than I died.
I felt the blackness come and go,
And strove to wake, but could not make
My sense climb up from below.
Do not believe that the gods cause illness and madness
and do not hope to bring healing and help the sufferer
by means of magic spells or false sorceries.
The Sacred Disease
The great contribution of Hippocratic medicine was its introduction
of an objective, naturalistic and empirical attitude to the biological--
and hence the behavioral--sciences. As early as 400 B.C., Hippocrates
had recognized the existence of epilepsy as a specific ailment (Esper,
1964, p. 119). Yet, even in Lord Byron's time--he himself was reputedly
an epileptic--seizures were still being attributed to mystical causes
rather than to any specific organic or functional disturbance. The
word seizure was originally used to describe epileptic convulsions
because the person was believed to have been seized by the devil.
Today these symptoms are known to be associated with abnormal
electrical activity in the brain. In some cases, this abnormal
activity can be shown to have its origin in a specific lesion in the
cortex. These cases are known as focal seizures. In many other cases,
no specific lesion or focus can be illustrated. These are nonfocal
epilepsies. The epileptic disorders have proven themselves to be stub-
born adversaries of modern medical science, and as yet there are no
generally reliable treatment modalities for the disabling and disturbing
handicaps epilepsy creates.
Recently, there has been interest in research which applies the
techniques of physiological feedback to the control of epileptic
seizures. While the results of this research are inconclusive, and the
entire effort has been fraught with controversy, the success that other
workers have had using physiological feedback techniques to treat
other clinical disturbances implies that this line of investigation
ought to be examined carefully. This review of the literature is
devoted to that end.
Definition of Epilepsy
It is important to note at the outset that epilepsy is not a
specific disease entity, but is rather the name given to certain signs
of central nervous system dysfunction. Penfield and Jasper (1954) have
described these clinical indications as a "state produced by an ab-
normal excessive discharge within the central nervous system" (p. 22).
Jasper, Ward and Pope (1969) define epilepsy as a "sudden paroxysmal
discharge" (p. 10) and add to it the concept of paroxysmal dysrhythmia
of the EEG and hypersynchrony to indicate not only excessive firing of
individual cells, but also massive discharge of many neurons in unison,
"abolishing the finely-organized tempero-spatial patterning character-
istics of the normal integrative activity of the brain" (p. 11).
Inherent in this definition is the diagnosis of a seizure disorder via
the abnormality of an individual's interictal EEG.
Epilepsy is often given as a diagnosis when clinical convulsions
and EEG manifestations are present, and the observed disturbances are
not secondary to an existing condition such as hypoglycemia, tumor, or
infection. Essential to this definition is a chronic recurrence of
Such a restriction is not entirely without reason, for although
the seizure is merely a sign of central nervous system dysfunction,
and thus has no one etiology or mechanism, the seizure in all cases does
utilize common neurophysiological mechanisms during its evolution. The usu-
al sequence of a seizure is characterized by aura, spread, generaliza-
tion, and cessation. The major motor seizure begins with an abnormal
electrical discharge in one part of the brain, spreads to other areas,
generalizing to a complete disruption of normal sensory and motor
activity, and then terminates (Jasper, 1969).
Medical Treatment of Epilepsy
While seizure activity may be the sign of some underlying neuro-
pathogic or neurochemical dysfunction of the brain, control of the
symptom alone is of immense benefit to the person with epilepsy and
indeed is the current goal of medical therapy. It is estimated that
approximately one-half of one percent of the population, or about
twenty million people worldwide, suffer from such seizures.
No single treatment procedure is effective in the treatment of
many epileptics. The condition is treated today principally through
medication with anticonvulsant drugs. These drugs, however, fail to
reduce seizure activity in 20 to 30 percent of all patients (Coatsworth,
1971). Even the patients who are helped by the medications risk side
effects that may be serious and debilitating. For example, phenytoin
sodium (Dilantin), perhaps the most widely prescribed antiepileptic
agent, is known to elicit some fifteen definite and eighteen possible
side effects (Bray, 1959), and to provoke a possible permanent degenera-
tion of cerebellar neurons (Selhourst, Kaufman and Horowitz, 1972;
Julien and Halperin, 1971). Phenacemide (Phenurone), a drug used in
particularly intractable cases, is highly toxic and can lead to such
bizarre psychic manifestations as suicidal ideation, paranoia, and
acute psychotic states (Livingston and Pauli, 1954).
Prior to the introduction of appropriate medication there was much
interest in dietary cures for epilepsy, but these treatments were not
effective. One dietary method was a dehydration regime pioneered by
Dr. Fay in Philadelphia in the 1930's. The ketogenic diet was developed
after investigators noted that seizures stopped shortly after the onset
of a fast, but the results did not extend beyond the starvation period.
One method involves starving and dehydrating the patient for a period
of several days, then prescribing a diet that contains four times as
much fat as carbohydrates and protein combined. The procedure works
best with children. This treatment can adversely affect the blood
vessels of adults and is extremely complicated to administer
Surgical intervention is indicated when medical treatment has
proven ineffective and the results of the operation on the patient's
brain will not alter functioning or cause debilitating symptoms more
severe than the seizures themselves. Surgical treatment is only
possible if the lesion is accessible and its removal will not severely
impair the functioning of the brain.
The forms of surgical intervention most commonly used are:
1. the excision of localized cortical areas exhibiting
constant abnormal electrical discharge;
2. temporal lobectomy for patients with clinical psycho-
motor seizures associated with focal, temporal lobe
electroencephalographic abnormalities; and
3. partial hemispherectomy in certain cases of epilepsy
associated with unilateral cerebral atrophy (Livingston,
Surgery clearly is indicated for only a small number of epileptics,
and the probability of application to the general population of
epileptics is limited. Rasmussen (1975) reported that of 129 patients
treated surgically for the removal of nontumoral lesions, 48 percent
has a moderate to no reduction of seizure frequency. Thus, none of the
conventional techniques for the treatment of epileptic disorders are
completely satisfactory. Additionally some 30 percent of the total
epileptic population is left without an effective therapy.
Learning-Theoretic Models of Epilepsy
The two main types of medical intervention described in the pre-
ceding section are all based on the assumption that the epilepsy is a
sign of some physiopathology of the central nervous system. And,
indeed, it is often the case that seizure activity is related to some
specific lesion. But the origin of nonfocal seizures, where no lesion
is known to exist, is a matter of controversy. While many tentative
hypotheses for nonfocal seizure activity postulate some form of neuro-
physiological abnormality, other investigators believe that en-
vironmental conditions external to the epileptogenic tissue may be a
major source of potentiating or attenuating the seizure disorder "
(Mostofsky and Balaschak, 1977, p. 723). The above authors provide an
extensive review of the behavioral studies dealing with the treatment
of epilepsy. They organize the review into several paradigms consist-
ing of: 1) reward management, 2) self management, and 3) phycho-
physiological procedures. Since physiological mechanisms provide a
more fundamental intervention procedure related to the possible central
mechanism to explain seizures, the remainder of this literature review
will concentrate on phychophysiological procedures. It is this type of
etiological thinking that provides a theoretical underpinning for the
use of physiological feedback as a therapy in cases of uncontrolled
epilepsy. The following section will examine theories of both types,
beginning with the physiological approach.
Learning Theory Models
Since physiological explanations are usually employed in the
explanation of epilepsy, several weaknesses occur when it is employed
as a unitary explanation:
1. how does the patient's condition differ on the one day
which he has a seizure as opposed to the five on which
he has none; and
2. what accounts for spread--can it be explained adequately
in physiological terms? (Rodin, 1968)
The origin of learning theory explanation of epilepsy was the
discovery in 1960 that a normal brain could "learn" to develop an
epileptogenic focus without any physical trauma to the cells (Mostofsky
and Balaschak, 1977).1 Morrell's work (1960) demonstrated the forma-
tion of a "mirror focus," or secondary epileptogenic lesion (SEL),in
the hemisphere opposite the existing lesion. By Morrell's definition,
a SEL is "an electrophysiologically defined area of paroxsymal discharge,
at least one synapse removed from the primary zone, but connected with
massive neural pathways" (pp. 538).
The mirror focus type of SEL that Morrell studied developed in two
phases appearing in the hemisphere opposite the primary lesion in the
homologous brain area. In the first stage, the mirror focus develops
spikes only in response to spiking in the primary focus. In the later
stage, the secondary focus becomes entirely independent, no longer
requiring the primary focus to act as a trigger.
The SEL can be viewed in relation to the "kindling phenomenon"
first reported by Goddard, McIntyre, and Leech (1969). Goddard under-
took his research to study the possible epileptogenic outcome of a
"chronic irritant" on cortical tissue. His theory arose from the
observation that "some patients who seem to have made good recovery
from known head injuries become epileptic at a later date" (p. 297).
Goddard's experiment involved the use of low level stimulation of
the brain of his experimental rats with implanted electrodes. He
found that subthreshold stimulation, inadequate to trigger a seizure
on its own, would eventually produce a seizure after repeated
1"Learning" is used loosely here to refer to changes in the
electrical characteristics of brain tissue in the absence of known
physical or chemical events preceding the change.
stimulation of the area. Subsequent stimulation would produce a
seizure after repeated stimulation of the area, even months later.
This phenomenon has also been observed by Racine, Gartner, and Burnham
(1972), Tress and Herberg (1972) and Wada (1976).
Why is a learning-theoretic model invoked to explain these re-
sults? Goddard observed in his experiment that the optimum frequency
at which the stimulation should be applied to produce an eventual
seizure in the rat was one a day. Stimulating the brain more often
deterred the development of eventual epileptic activity. It is hard to
understand this in terms of direct tissue damage; moreover, the magni-
tude of the potential applied to the brain as a stimulus does not
effect the time interval before the area becomes an epileptic focus.
The neurons appear to exhibit a type of learning curve, reaching
100 percent accuracy when the area becomes an autonomous epileptic
focus. The kindling phenomenon has also been observed to occur when
subthreshold doses of seizure-inducing drugs such as pentylenetetrazol
(Metrazol) are administered.
The learning-theoretic model offers an explanation for the develop-
ment and spread of seizure activity. The immediate question, then, is
whether the process can be made to work in reverse, that is, can the
brain learn regulatory or pacemaker activity? To answer this question
a discussion of normal brain wave patterns and their controlling
mechanisms will be presented. Based upon these concepts, the possi-
bility of applying EEG physiological feedback as a means of attenuat-
ing the evolution of paroxysmal activity will be explored.
One way of analyzing EEG activity is to divide it between
synchronous and asynchronous patterns. Synchronous patterns consist
of steady, periodic activity which ultimately relate to the extent to
which neuronal activity is simultaneously active. In EEGs it has been
characterized by the coincidence of waves of similar shape. Though
there have been no exact quantitative measurements of synchrony and
its identification is usually based upon experience, it has been
estimated that normal synchronous activity requires between 0.1 to
1.0 percent of a cell population firing to produce a lOOuV wave on the
surface (Diver and MacGillivary, 1977). This activity has been described
by Chase and Harper (1971) as essentially cortical idling--the absence
of any specific information processing activity. Normal synchrony has
been most notably associated with alpha and slow-wave sleep.
Probably the best studied of human synchronous rhythms is the
alpha rhythm (8-12Hz), whose role in relaxation processes has been
lately popularized. The alpha frequency was discovered in 1929 by the
German neurologist Hans Berger (Berger, 1929). He learned of the
inhibitory nature of the alpha rhythm by observing that its production
was blocked by directing attention toward a stimulus or event. Adrian
and Matthews (1934) found that alpha is not always blocked by any form
of attention; what is required is the perception of attempted per-
ception of a pattern. They also found the difficulty of the task to
which attention was directed was a factor in the degree of alpha
blocking that occurred. Modern research on human occipital alpha has
shown that this pattern is likely to be dominant when the oculomotor
and visual systems are not activity tracking a target or attending to
Asynchronous patterns appear as low voltage fast waves, similar
to random electrical activity and are associated with active attention
and the processing of information. These patterns in the frequency
range above 13 Hz are identified as beta activity (Kiloh and Osselton,
Inhibitory Synchronous Activity and Seizure Production
Andersen and Andersson (1968) postulate that the cortex is almost
totally dependent on the thalamus for its synchronous activity. Ac-
cording to their Facilitative Pacemaker Theory, all major thalamic
nuclei have the capacity to generate rhythmic activity. Each nucleus
functions as a small pacemaker, projecting to a specific cortical
area. There may be as many as thirty to forty thousand of these
pacemakers. Diver and MacGillivary (1977) describe the general principle
underlying pacemaker activity in the generation of synchronous rhythms,
particularly the alpha rhythm. They state that inhibitory activity
is promoted by a series of feedback mechanisms that cause a synchronous
inhibition through the action of a class of neurons, which they term
recurrent inhibition. Eccles (1969) has characterized inhibition as
"that action by a class of synapses that opposes excitation and tends
to prevent the generation of impulses by excitatory synapses" (p. 235).
Thatcher and John (1977), in speaking about inhibitory synchronous
activity, posit an interactive loop process which keeps excitatory/
inhibitory activity in bounds, and which functions to "prevent epilepsy
and regulate the levels of excitation such as sleep and wakefulness"
(p. 117). Thatcher and Purpura (1972) further argue that thalamocorti-
cal synchronization, particularly as it applies to the alpha rhythm,
may be characterized as a "scanning mechanism" which alternately excites
and inhibits cortical activity. According to Thatcher and John (1977)
a subcortical scanning pulse exists that monitors the state of neural
excitability. They cite evoked potential research which demonstrated
a phasic alteration in excitability corresponding to the frequency and
phase of the alpha rhythm. During faster, desynchronous activity,
usually coincident with active focusing of attention, the loop systems
are broken down into a large number of smaller loops or independent
oscillators parsing activity according to the specific system involved.
Hypersynchrony, considered by most the hallmark of seizure pro-
duction, has been characterized by Jasper (1969) and Penfield and
Jasper (1954) as having two main functional components: 1) existence
of an excessive firing of individual cells and 2) the excessive
synchronization of unit discharge activity which does away with the
organized patterned electrical activity that is characteristic of
normal brain function.
In discussing the mechanism of synchronization, Diver and
MacGillivary (1977) state that the production of normal alpha activity is
far different from the hypersynchronous patterns associated with
seizures. The general lack of inhibitory mechanisms in the recruit-
ment of adjacent neuronal areas is considered the prime mechanism of
the genesis of hypersynchrony. If the interactive loop hypothesis of
Thatcher and John is invoked, the excitatory/recruitment phase is not
counterbalanced by the normal inhibitory one. In such a case,greater
and greater portions of neuronal activity begin to respond without
Given the consensus that seizure activity is characterized by a
lack of inhibitory synchronous activity, and in view of the concepts
derived from the SEL and the kindling phenomenon, the technique of
physiological feedback may be an efficacious method of establishing an
inhibitory augmentation mechanism. The remainder of this review will
be aimed at further development of this concept, beginning with an
overview of the topic area.
Physiological feedback is a technique wherein the subject is given
continuous information about the state of some physiological variable of
which he is not normally aware, such as blood pressure or EEG activity.
As currently used, the technique involves the amplification of
minute biological electrical potentials into signals large enough to
record on a polygraph. The signals are then filtered to detect a cer-
tain frequency and voltage band to distinguish changes of a predeter-
mined sort. The information is then fed back to the subject using
devices such as tone generators, slide projectors, or indicator lamps.
In this manner, the individual is given information about his physio-
logical state in a useful form. Individuals have been taught with this
method to exhibit a heightened degree of control over heart rate, blood
pressure, galvanic skin response, muscle tension, skin temperature, and
EEG alpha activity (Barber, DiCara, Kamiya, Miller, Shapiro and Stoyva,
1970-77; Miller, 1978).
Researchers have investigated the applicability of physiological
feedback as a therapeutic intervention into a great variety of patho-
logical states. Some conditions for which physiological feedback
therapy seemed to yield promising results were muscle retraining after
atrophy or paralysis (Andrews, 1964), chronic anxiety (Raskin, Johnson,
and Rondestvedt, 1973), tension headache (Epstein, Hersen, and Hemphill,
1974; Budzinski, Stoyva, and Adler, 1970), heart rate control in
patients with heart disease (Troyer, Twentyman, Gatchel and Lang, 1973;
Weiss and Engle, 1971), and control of essential hypertension (Benson,
Shapiro, Tursky and Schwartz, 1971).
In these studies various physiological measurements were used as
the variables that the subjects were taught to control. In the studies
on muscle retraining, anxiety, and tension headache, the variable was
muscle tension as recorded on an electromyograph. For the insomnia
study, persons were taught to augment the theta wave of the EEG. In
heart rate and blood pressure studies, the subjects were taught to
control those variables directly.
While a variety of work has been done in the treatment of clinical
disorders through physiological feedback techniques, only recently has
EEG physiological feedback been used to investigate pathological
states (Mulholland and Benson, 1971). Prior to this time, most
physiological work using EEG rhythms as feedback variables was devoted
to studying normal processes, such as sleep, meditation, and relaxation
(Jonas, 1973). There is a growing body of literature reporting attempts
to treat epilepsy through physiological feedback techniques.
The Initial Use of Physiological Feedback in the
Treatment of Epilepsy
Early experiments in conditioning approaches to the treatment of
epilepsy differed from the methods of physiological feedback. In one
early sample, Stevens (1962) and her colleagues (Stevens, Milstein, and
Dodds, 1967) attempted to extinguish spiking and other abnormal
epileptiform EEG patterns. A tone and mild electric shock were
administered to the subject at the moment the abnormal wave form
occurred. When the tone was discontinued, discrimination disappeared
in all but one subject. That person was still able to respond in the
absence of the tone when there was a paroxysmal discharge.
Another model for learning-theoretic intervention was
explored by Efron (1956, 1957) who utilized the model of classical
conditioning. His work was dependent upon the existence of an uncon-
ditioned stimulus, an odor, in the case he studied, which could arrest
the seizure activity of the patient. He was then able to condition a
visual stimulus, a bracelet, to the dissolution of the aura.
In a series of studies, Forster and Booker (1964) and Foster
(1967 and 1972) developed a method for treating epilepsy where the
seizures are induced by sensory stimuli. A stimulus which usually
evokes a seizure was used via an extinction or habituation procedure
which is designed to weaken the ability of the stimulus to provoke a
seizure. Mostofsky and Balaschak (1977) have suggested that rather
than an extinction or habituation paradigm for the attenuation of
sensory specific induced seizures, Forster's method tends more to
shift attention away from the seizure stimuli and, therefore, act as
a competing stimuli that does not evoke the paroxysmal response.
The groundwork for efforts using the principles of operant con-
ditioning in the treatment of seizures was laid by Sterman, LoPresti,
and Fairchild (1969) who reported that cats who had been taught to
produce sensory-motor rhythm (SMR), which ranges in frequency between
12-16 Hz measured over the coronal gyrus, were unusually resistant to
seizure-producing drugs. In cats trained to increase SMR activity,
there was a significant delay between the administration of the drug
pentylenetetrazol (Metrazol) and the onset of preseizure symptoms. In
untrained cats, no such delay was noted; the administration of the drug
was followed by a much shortened interval between preseizure symptoms
and the seizure itself.
A few years later, Sterman and a colleague produced the first
report of a treatment of a human epileptic through this technique
(Sterman and Friar, 1972). In this case they reinforced an 11-13 Hz
rhythm, which they maintained was the analog of feline SMR. This
represented a break with earlier research that considered the mu, or
en arceau rhythm (7-11 Hz) to be the human correlate of the sensorimotor
Sterman and Friar (1972) now claim that human SMR is a rhythm
distinct from both the alpha and mu rhythms. The frequency that Sterman
and his colleagues have been identifying as SMR is extremely difficult
to detect, due to its low amplitude. While most cortical synchronous
rhythms are in the range of 50 PV or larger, human SMR is approximately
5-15 PV. It has thus been difficult to verify the existence and
properties of this rhythm in the human EEG.
Kaplan (1974) reported that the low amplitude SMR was discernable
in human EEG's through the use of a dual filtering system. The signal
was prefiltered through a conventional analog filter and then digitally
filtered on a PDP-12 computer.
Kaplan (1974) pointed out that the system used by Sterman and
associates was unable to distinguish SMR accurately and, therefore,
there was no proof that it was indeed SMR that was being reinforced.
She was especially critical of the fact that in none of Sterman's
experiments was there any documentation that the levels of SMR were
increased by the feedback training sessions. In her experiment there
could be no doubt about which rhythms were being reinforced.
Three dependent variables were measured in the Kaplan (1974)
study. They were evaluations of clinical EEG's, seizure incidence,
and power spectra (fast Fourier transforms of EEG epochs). The experi-
ment involved the training of two epileptic patients with 12-14 Hz
biofeedback for three months. No decrease in seizure activity re-
sulted, and one of the patients dropped out of the study. The
remaining patient and two others were then trained in the production
of the alpha frequency.
Necessary changes in the medication that the one patient in both
stages of the experiment received made it impossible to compare that
subject's results directly. It is believed that the observed changes
in both seizure incidence and power spectra for this subject are due to
the medication changes.
Kaplan's other two patients, trained only in the theta-alpha
production (6-12 Hz), showed a significant reduction in seizure
incidence. However, no changes in the power spectra of the EEG's were
evident. Thus, the clinical changes were not ascribed to learning of
EEG synchrony. Rather, Kaplan attributed the changes in the subjects'
learning to function on a lowered level of arousal. That two of the
subjects later reported an increase of seizure incidence under stress
was taken as a further indication of this interpretation.
In a later paper, Kaplan (1975) reiterated her criticisms of the
methodology employed by Sterman. She emphasized in particular the
impossibility of ascribing the observed clinical changes to changes in
the synchronous rhythm production if the fact that these rhythms were
indeed being produced was not demonstrable.
Gastaut (1975), the discoverer of the mu rhythm, agreed with
Kaplan's criticisms and maintained that there was in fact no rigorous
evidence that biofeedback training was ever effective in treating
In the last few years, several researchers have attempted to
answer some of the questions and controversies surrounding this
earlier research. Table 1 presents a summary of selected EEGphysiological
feedback studies in epilepsy. This more recent research has tested
the effectiveness of alpha and theta rhythm training as well as the
human SMR and has experimented as well with the reinforcement of
combinations of synchronous activities of different frequencies. The
results of this research are also equivocal and leave the issue far
Cabral and Scott (1976) contrasted the effects of alpha training
with a Jacobsonian relaxation technique on seizure reduction. Their
conclusion was that alpha training had a more consistent effect on
Kuhlman and Allison (1977) designed a study in which EEG train-
ing was used to effect significant seizure reduction in three out of
five patients. The study was designed to permit the determination of
the actual changes in the EEG produced. Prior to the initiation of
training to reduce seizures, three normal subjects were successfully
trained to enhance the central mu rhythm in the 9-11 Hz band using
50 20-minute sessions. Similarly, three normal subjects given the
same amount of training were unable to increase 12-14 Hz activity with
Table 1. Summary of EEG Feedback Studies in Epilepsy.
Study Frequency Subjects Design and Duration
Sterman & Friar
Seifert & Lubar
Lubar & Bahler
& Inch (1976)
11-13 Hz (+)
12-14 Hz (+)
11-13 Hz (+)
12-14 Hz (+)
4- 7 Hz (-)
12-14 Hz (+)
6-12 Hz (+)
9-14 Hz (+)
9-14 Hz (+)
6- 9 Hz
6- 9 Hz
feedback training. When training to attenuate seizures was instituted,
none of the successful patients increased activity in the 12-14 Hz
range but did enhance significantly production in the 9-11 Hz range.
This same study observed that occipital alpha increased in the
successful cases. They state that "in fact, the increase in the
occipital alpha frequency is a virtual mirror image of the seizure
frequency" (p. 122).
One further implication of these results is that seizure reduction
is not dependent on alteration of a specific EEG pattern, or the increase
of any specific inhibitory cortical rhythms. Research byQuy (1977)on
three chronic seizure patients led him to the conclusion that indeed the
most effective form of biofeedback training for these patients was one
which reinforced all synchronous activity in a broad, midrange spectrum.
A number of investigations have borne out this role of the mid-
range spectrum in seizure reduction. In particular, the suppression
of the slow theta frequencies as well as the enhancement of the midrange
spectrum seems to bolster the effectiveness of the treatment. This
result has been reported by Finley, Smith and Etherton (1975), Lubar
and Bahler (1976), Wyler, Lockard, and Inch (1976), and Sterman and
Budzinski and Stoyva (1969) reported that when frontal EMG levels
became particularly low, the subjects tended to display an increase in
EEG theta waves. Sittenfeld, Budzinski, and Stoyva (1976) demon-
strated that profound muscular relaxation was associated with the
appearance of theta in the EEG. Budzinski, in "Clinical Implications of
Electromyographic Training" (1977, p. 438), reported that at the lower
levels of frontalis muscle tension there arose an an inverse relationship
with EEG theta.
Beatty, Greenberg, Deibler, and O'Hanlon (1974) reported that the
absence of theta (which they defined as 3-7 Hz) was particularly as-
sociated with the ability to maintain vigilance. O'Hanlon and Beatty
(1975) showed that theta suppression may prevent or lessen the perform-
ance decrements typically associated with long, exacting vigilance
The link between theta rhythm and epileptic seizures has been
firmly established. Gibbs, Gibbs, and Lennox (1937), in a study of
400 epileptic patients, noted the tendency of seizures to occur when
the patient is awakening or falling asleep. Theybelieve pre- and post-
sleep stages modify the rate of normal cortical rhythm, placing a
strain on the self-regulatory mechanism of epileptics with which they
are unable to cope.
Gibbs, Fuster, and Gibbs (1948) and Jasper, Pertuiset, and
Flanigin (1951) have confirmed the focal cortical origin of theta
rhythms in a majority of epileptic patients.
In 1974, Johnson and Meyer devised a phased biofeedback treatment
procedure for an eighteen-year-old female epileptic who was allergic
to Dilantin. The treatment was administered in a phased sequence
beginning with alpha feedback, followed by alpha and theta feedback,
and concluding with theta feedback alone, in an attempt to establish
a low arousal antistress response. The patient, though unable to stop
a seizure once begun, was able to prevent the onset of seizures by
going into an alpha state when she felt an aura developing. At the
end of one year she had experienced a 46 percent decrease in seizures.
Sterman's most recent study (1978) was designed specifically to
investigate the importance of specific training frequencies in the
interpretation of EEG data. In his double blind study,patients were
alternately trained to produce EEG activity in one of three frequency
ranges. One was always the 6-9 Hz theta frequency, and the other was
either 12-15 Hz or 18-23 Hz. Patients were rewarded for producing one
in the absence of the other. After three months the contingencies were
reversed without the patient's knowledge. For six of the eight sub-
jects, significant seizure reduction occurred when the higher frequency
was trained and the lower suppressed. When the higher frequency was
within the 12-18 Hz band, the beneficial effects survived the reversal.
For the six patients with seizure reduction, the average reduction in
rate of seizures was 74 percent. Sterman interprets this as illustrat-
ing the specific role of training for various frequencies and rejects
the claims by some detractors that the beneficial effects are an
artifact of the design, produced by the patient sitting motionless and
functioning at a lower level of arousal.
Summary and Hypotheses
Epilepsy is the term applied to a complex set of central nervous
system dysfunctions with electrical and most often behavioral con-
sequences. A further delineation of this syndrome would characterize
it as an excessive firing of many neurons in unison, hypersynchrony,
and slow wave activity, particularly in the theta range. This has
been considered the hallmark of seizure initiation and propagation.
Complex partial seizure activity is associated with abnormal EEG ac-
tivity which generally begins in a localized area and generalizes as
the seizure develops.
As conventional therapies for seizures of this sort are unable
to effect significant improvement in 20-30 percent of epileptics,
interest in the possibility of alternative treatment modes by elicit-
ing learned changes in the brain's electrical activity became more
This interest was spurred by observations, which seemed to
indicate that learning processes might,in some cases, be involved in
the development of seizure activity. Morrell's discovery of the mirror
focus phenomenon and Goddard's observation of the kindling phenomenon
are the two principal pieces of research in this area.
The techniques of biofeedback, which have been used with success
in the treatment of a great variety of other clinical entities, were
used to train patients in the production of these cortical rhythms.
Initially a great deal of interest was focused on those frequen-
cies believed to be the human analog of the SMR in cats. Feline SMR
was known to be associated with the suppression of motion and was also
observed to be correlated with resistance to seizures.
Further research has cast doubts on the specificity of the SMR
as a factor in seizure reduction. Criticisms of some of these early
studies, especially by Sterman et al. (1974), pointed out the importance
of illustrating changes in the EEG as well as proving seizure reduction
if any notion of cause and effect is to be developed.
It is the learning theory of seizure propagation that leads to
the physiological feedback method of therapy. There is considerable
evidence to suggest that brain cells can learn to be epileptogenic,
either through mirror focus mechanisms, Morrell's secondary epilepto-
genic lesion, or through the Goddard kindling phenomenon. If the
brain can learn to be more easily excitable, can it conversely be
made to learn inhibitory or pacemaker activity? This is what physiological
feedback techniques attempt to accomplish.
In order to be effective, however, considerably more information
on the precise mechanisms involved must be developed through research.
Even if one accepts the learning theory, what waveforms should be
reinforced? Sterman and his colleagues use a waveform assumed to be
the human counterpart of the cat SMR and have claimed success with a
small patient population. Additional study, however, has convinced
the same researchers that SMR in humans exists as a rhythm distinct
from either alpha or mu. Kaplan assumed that previous experiments were
using a filter system which could not discriminate signals in the ap-
propriate range adequately. Using more sophisticated electronic equip-
ment, she established that the Sterman rhythm, when used in biofeed-
back training, did not work. Using low frequency synchrony was also
The literature,therefore,appears to indicate that biofeedback
techniques are promising avenues to explore in the treatment of
epilepsy, yet considerable fundamental research is needed before
understanding and effective therapy can be accomplished. The litera-
ture dealing with seizure initiation and propagation points to
inhibitory synchronous rhythms, especially the alpha rhythm, as a
means of attenuating this activity. Enhancing the percentage of time
the patient produces alpha while simultaneously diminishing the effects
of hypersynchrony via learned suppression of slow wave theta may well
provide a direct and parsimonious means of controlling seizure
The hypotheses to be tested are as follows:
1. There will be a significant difference between pre- and post-
EEG measurements. This difference will be manifested by a diminution
of slow wave activity, chiefly in the theta (3-8 Hz) and delta (1-3 Hz)
range, and an increase in the percentage of time in the faster range,
particularly the alpha (9-13 Hz) range.
2. Seizure incidence will diminish over time, as a function of
training, being significantly less at the completion of the study.
3. The time series measuring the rate of alpha production during
and following intervention can be represented by a significant (positive)
change in both level and slope as compared with baseline measures.
4. The time series measuring the rate of theta suppression
during and following intervention can be represented by a significant
negative change in level and slope as compared with baseline measures.
The basic operations required to investigate these hypotheses
1) measures of concurrent dependent variables of alpha percent,
average frequency, and average amplitude, theta percent and theta
count for each subject over a sufficiently long time course to permit
an evaluation of baseline, treatment, noncontingent treatment;
2) a measure of seizure activity via a diary begun before treat-
ment and kept throughout; and
3) an independent measure of EEG frequency changes pre- and post-
Six subjects were selected using the following three main
criteria: seizure history, frequency of seizures, and subject's being
refractory to anticonvulsant medication. First, it was required that
each subject have a positive history of essentially uncontrolled
seizures for at least five years prior to the initiation of training.
They were identified via patient records at the Veterans Medical
Center, Gainesville, Florida, and through referrals from the Jacksonville
Epilepsy Foundation and subsequently through medical records of the
patient's neurologist. The second requirement was that following
long-term anticonvulsant therapy with good compliance, the seizure
condition was not under adequate control. Finally, the requirement
was established that the subject must be experiencing at least five
seizures per month that were verified by family members, friends, or
Beginning approximately one month prior to the initiation of
training and continuing throughout training, each subject was sup-
plied with a seizure diary on which he indicated by day and time the
occurrence of seizures. In each case family members or close friends
were asked to monitor this record to ensure its accuracy. Subjects
were frequently reminded of the importance of this diary and asked to
bring it in weekly to assure understanding and compliance.
The following section presents Individual Subject Profiles
indicating summaries of each subject's background medications, EEG
data and pertinent medical and behavioral information.
Individual Subject Profiles
Subject #1 was a fifty-four-year-old divorced white male with a
history of seizures from age thirty-one. He has been diagnosed as
having complex-partial seizures with a right temporal parietal focus.
The patient has been operated on with right parietal lobe surgery as
a result of an old trauma. Medication prior to and throughout train-
ing consisted of Dilantin 300 mg.a day.
A review of EEG records revealed an asymmetry of fast activity
because of high amplitude over the right parietal and temporal regions.
Also, continuously low voltage and irregular slow waves in the theta
and delta range were seen over the right parietal and temporal regions.
This was considered an abnormal EEG because of the above. These
alterations were considered compatible with a structural lesion. The
asymmetry and lateralized fast activity were thought to be due to the
The patient's typical seizure pattern consisted of tactile
sensory loss accompanying partial hand and facial paralysis on left
side and an expressive aphasia ranging frommarkedly slurred speech to
no speech at all. There was no loss of consciousness during these
Seizure frequency at the beginning of the study: 59 per month.
Subject #2 was a twenty-three-year-old single white male with
history of seizures from age ten. He has been diagnosed as having
psychomotor seizures with an unknown etiology. Medications prior to
and throughout the course of training consisted of: Tegretol, 600 mg.
a day; Tranxene, 7.5 mg.,five times a day.
A review of the patient's history indicated some behavioral
problems as a child, angry episodes, acting-out. These behaviors
have not been noted in the patient's record since 1971.
Review of EEG records revealed an initial abnormal EEG with left-
sided predominance; sharp activity was seen over both midtemporal and
posterior temporal areas. Later records indicated diffuse bitemporal
slowing with background rhythms almost continuously irregular in the
2-1/2 to 5 Hz range over the temporal, parietal, and occipital
regions. The latest clinical interpretation commented that little if
any trace of alpha rhythm was evident. The patient's diagnosis has
uniformly been psychomotor seizure presently with a bitemporal focus.
The patient experiences an aura with a "scary" feeling beginning
in his stomach and feeling of pressure in his head. A typical seizure
includes lip smacking and finger movements. There is a post-
ictal stupor and tiredness that lasts from several minutes to
Seizure frequency at beginning of study: 27 per month.
Subject #3 was a thirty-two-year-old white married male with a
history of seizures, post traumatic in etiology, since 1975. Medica-
tions at the outset of training consisted of Dilantin 400 mg. a day
and Tegretol 1000 mg. a day. During the course of training Celontin
500 mg. was added.
Review of EEG records revealed an abnormal EEG with focus of
epileptiform discharge in the right anterior temporal regions.
Throughout the tracing a focus of epileptiform sharp waves and spikes
and slow waves were identified emanating from the right anterior
Behaviorally the patient has been described by his neurologist as
having a decreased libido, a seemingly growing disinterest in sexual
relations with his wife though no reported inability to achieve and
maintain an erection, and a significant weight loss. The patient has
been described as being depressed due to his inability to maintain
his former lifestyle as worker and provider.
A typical seizure for this patient manifested itself with amnesia
and automatic auras of short and mild duration as to be not useful in
self-control. He also manifested various kinds of situational automa-
tisms. This subject attended the study for three baseline days and five
days of treatment trials but decided to drop out of the experiment.
Seizure frequency at beginning of study: 20 per month.
Subject #4 was a twenty-four-year-old white divorced female
with a history of seizures since age nineteen. She has been described
as having psychomotor seizures of traumatic origin. The patient
suffered a closed head injury in an auto accident two years prior to
the onset of seizures. Medication at the outset and during the course
of training consisted of Dilantin 100 mg.,five times a day.
Review of EEG records revealed an abnormal EEG with epileptiform
discharges in the right posterior temporal region. There was signifi-
cant slowing into the theta and delta range over these areas and the
Behaviorally the subject presented herself as the one patient
most positive and interested in maintaining a normal lifestyle which
included work and family. Even when the frequency of her seizures
made it impossible to work, she continued to plan for the time when
she would again be able to work. Despite large difficulties in her
personal life, the patient was able to complete training with few
A typical seizure for this subject consisted ofa feeling of
paralysis, a severe memory disruption, and often a feeling of dis-
orientation and anger postictally.
Seizure frequency at beginning of study: 207 per month.
Subject #5 was a twenty-six-year-old white married male who was
injured and sustained a subdural hematoma in 1969. The onset of
seizures was in 1971 with frequent major convulsions and brief
complex partial automatisms. Medication prior to and throughout the
course of training were Mysoline 1000 mg.a day and Dilantin 300 mg.
A review of EEG records revealed an initial abnormal EEG with
the primary focus over the right temporal area. Subsequent recordings
indicated a markedly abnormal tracing with the occurrence of bi-
synchronous generalized epileptiform activity maximal over the right
hemisphere. The patient's present diagnosis is a complex partial
seizure disorder with a right temporal focus with a secondary bi-
lateral synchrony. There is some generalization of seizures to the
grand mal type.
Seizure frequency at beginning of study: 19 per month.
Subject #6 was a twenty-four-year-old white single male with a
history of seizures from age twelve. He has been diagnosed as having
psychomotor seizures with a left-temporal focus secondary to two
surgical procedures in the left fronto-temporal area for obliteration
of an AV malformation in the middle cerebral circuit. Medications
prior to and throughout the course of training consisted of Dilantin
500 mg.a day and Celonten 1200 mg.a day.
A review of the patient's history indicated a psychiatric hospi-
talization for an acute psychotic episode. Presently he is attending
a day hospital program and is well motivated to seek employment.
A review of EEG records revealed an abnormal EEG with a dis-
organized rhythm with voltage in the range of 6-8 Hz. There was
considerable slowing in the delta range with a discrete spike focus in
the left midtemporal area (seen over the T3 electrode) and considerable
theta activity also. The diagnosis was a psychomotor seizure condi-
tion with a primary left-temporal focus.
The patient experienced an aura, a peculiar feeling in his chest,
and a salty taste in his mouth. His typical seizure was like a dream
of falling, temporary expressive aphasia and loss of train of thought.
Seizure frequency at beginning of study: 147 per month.
Prior to the initiation of training each patient was administered
an EEG using bipolar referents T3-T4, C3-C4 with an ear ground (using
the 10-20 electrode placement system). The C3-C4 placement was collected
on channel 1 and T3-T4 was collected on channel 2. Twenty to fifty
minutes of recording was obtained for each subject. They were in-
structed to stay awake throughout the recording sessions but asked to
keep eyes closed. Though every attempt was made to instruct subjects
to remain awake, one subject, #1, did fall asleep during the post
session. The light sleep portion of his record was omitted from the
analysis. The record was obtained on paper and FM tape. This pro-
cedure was again followed immediately upon the completion of training.
During the preparatory phase each patient was individually inter-
viewed using Mostofsky and Balaschak's (1977) seizure questionnaire and
given a copy of a short article entitled An Introduction to Biofeedback
(Fuller, G. and Sempell, P., 1977), a layman's guide to physiological
feedback. This article was the basis for subsequent discussions of
the nature of physiological feedback. A tape recording instructing the
patient in proper procedure and providing samples of the kind of feed-
back sounds they might expect was played four times (see Appendix A for
script)--once during the initial interview and prior to sessions 4-5
and 6--so as to familiarize them with this feedback mode. It was
constructed to duplicate auditory feedback of 20 percent, 50 percent,
and 100 percent success in producing alpha. They also were provided
feedback for theta (alarm) and that of artifact from too much movement.
Subjects were instructed to increase the occurrence of one tone and
suppress the alarm. All subjects were instructed to keep their eyes
closed. On a chalkboard in the feedback room a diagram showing the
various frequencies and their respective feedback was provided to enable
the patient to grasp more concretely the nature and rationale behind
the desired response.
Training took place from three to four times per week with a ses-
sion, including electrode placement and posttraining feedback, lasting
one and one-half hours. Standard Grass Disc Electrodes were used for
recording. Placement was bipolar, using T3-T4 and C3-C4 for alpha en-
hancement and theta suppression, respectively, and an ear ground.
Hewlett-Packard Redux electrolyte paste was employed and the electrodes
were cemented to the scalp with collodion and dried with a compressed
air stream. The criterion for acceptable electrode contact was less
than 10,000 ohms resistance as measured on a digital volt-ohm meter.
Patients were seated in a comfortable, reclining chair, in a
quiet room 15 x 13 feet. The room was further subdivided and darkened
via the use of two cushioned standing partitions that served to
militate against any extraneous light and sound and give the feedback
space an isolated smaller appearance. To the rear and approximately
three feet away sat the experimenter who monitored all patient
behaviors and all feedback through a common headphone setup.
Feedback consisted of an auditory tone that varied in pitch (sig-
nifying frequency changes in the pass band) and loudness (indicating
EEG amplitude within the pass band) whenever the subject produced six
successive zero crossings in the alpha range (9-13 Hz). Also there
was an alarm that sounded whenever his EEG frequency exhibited six
successive zero crossings in the theta range (3-8 Hz) and remained on
until such time that he ceased to produce waves in this range.
Using an A-B1-C-B2(D) design, each patient was run through
between 32 and 36 fifty-minute trials. Figure 1 indicates
PreTraining PreTreatment Treatment 1 NonContingent Treatment2 Theta EEG
EEG Baseline Phase Suppression Clinical
aver. freq.& SD
aver. amp.& SD
3 sessions 18 sessions 4 sessions 6 sessions 3-4 sessions
Schematic representing sequential operations of treatment
the various operations and measurements over the entire study for
each subject. Each trial was divided into ten five-minute segments
(see Figure 2) with a 45-second lag time when no data were collected.
This lag time enabled the alpha-numeric component of the 5400 system
to print out the previous segments data. At the beginning of this
intermission the patient was required to respond to a light touch on
the arm by lifting the index finger on the right hand. The patient
was provided with the digital feedback obtained consisting of alpha
wave percentage and theta wave percentage. Supplemental encouragement
was given to enhance the alpha level and suppress the theta by keeping
the sound on and the alarm off.
The initial three trials (Condition A) were baseline sessions
when no feedback was provided. From trial four through twenty-two
(Condition Bl),contingent feedback was provided for segments two
through nine. Segments one and ten were again baseline where the
subject was not provided with auditory feedback. Sessions 23-26
(Condition C) consisted of the identical ten-segment procedure but
utilized noncontingent feedback for the eight feedback segments.
Tape recordings of each subject's sessions 21 and 22 were interfaced
with the feedback apparatus. Sessions 23 and 25 utilized the tape
recording from session 21 and sessions 24 and 26 were based upon
session 22. All procedures were identical to contingent trials
including the usage of the digital feedback from these previous
sessions. Upon debriefing at the conclusion of the study,it was
clear that none of the patients were aware of this manipulation as the
tape recorder had been used various times in the past during sessions
to accustom the patients to its presence.
- Treatment phase -
8 5 minute segments
tone for alpha
alarm for theta
Wiri"-Ir v I -i
10 5-MINUTE SEGMENTS
Schematic representing typical treatment trial with ten
I ---L9(~ __ _c_-~ .1 __
Sessions 27-32 (Condition B2) were again contingent trials when
feedback was provided depending upon the subject's alpha wave produc-
tion and theta suppression.
For three of the subjects (#2, #5, and #6) C3-C4 location was
employed as the active locus for feedback (Condition D), and the
subjects began receiving the analogue tone indicating only theta's
presence. They were told that the presence of the tone signaled theta
and were instructed to keep the sound off. Subjects #2 and #5
received four such trials while #6 received three theta suppression
The apparatus consisted of several major components: two EEG
waveform analyzers (an Autogen 120a and Autogen 70a), an Autogen 5400
data acquisition unit, and a tape recorder. Both the 120a and 70a
operate as variable bandpass filters for the EEG within the range
from 2-20 Hz. The Autogen 120a permits feedback for EEG frequency and
voltage levels continuous within its range of 2-20 Hz and 10-150 uV.
The 120a was set at 100 PV so as not to reinforce epileptiform EEG
activity. The output of the Autogen 120a consists of frequency,
amplitude, percent time in a pass band, and the output of the 70a of
discrete occurrences of a selected frequency within a pass band from
which both a percent time and a count measure was obtained. The 120a
was used to measure alpha (9-13 Hz) along the three parameters
specified above and provide an auditory feedback, varying in pitch
and loudness when the patient was producing EEG frequencies within
this range. Pitch varied with EEG within the pass band while loudness
varied with EEG amplitude shifts. A distinctive alarm sounded when
the subject's EEG frequency fell 1 Hz or more below the lowest fre-
quency in the pass band and came on at 8 Hz. This alarm served two
purposes: 1) to warn the subject of potential drowsiness and 2) to
serve as a theta indicator. The Autogen 70a was used to record the
percent time of theta and the number of discrete occurrences in the
five-minute period. The Autogen 120a monitored from T3-T4 leads
while the 70a monitored from C3-C4 leads. This was reversed only in
the last condition (D).
The Autogen 5400 and the attached Alpha-Numeric Printer were
employed to tabulate the data from the 120a and 70a and produce a
paper tape recording that was timed to print out a data summary every
five minutes. In addition to the means of each of the variables it
provided standard deviations of alpha amplitude and average frequency
standard deviations between 2-20 Hz.
The Akai tape recorder was interfaced with the feedback units to
allow for tape recording and playback capabilities during the
The two channels of EEG data collected were obtained, before and
after physiological feedback training, both on chart paper and FM
tape, C3-C4, channel 1, and T3-T4, channel 2, (International 10-20
electrode system). They were analyzed via the method of frequency
analysis. The program, run on a PDP-8E computer, was designed to
analyze results between the frequencies 1-20 Hz in 0.5 Hz increments.
The input from the FM tape was prefiltered with a Krohn-Hite filter set
between 1-20 Hz to exclude frequency artifact. The program sampled the
input 1000 times/second and timed each zero crossing, storing each in
a register. At the end of a ten-minute epoch, input was terminated,
and the contents of each register were printed. It then measured the
time elapsed since the last data input and formed a histogram indicat-
ing the number of counts in each of the 0.5 Hz period bins. In addi-
tion, the average frequency within each of seven bands (1-3 Hz,
3.01-5.99....18.1-20 Hz) together with the percentage of total
time within each of the pass bands, was tabulated and printed. The
values of these bands were selected to approximate the ranges of
delta, theta, and alpha frequencies in three Hertz increments. To
produce a meaningful frequency axis, the data were collected in time
intervals such that their inverse was close to the integer frequencies
and their midpoints (i.e., 1, 1.5, 2, 2.5, etc.).
A Chi-square test was performed to examine the two histograms
to test the hypothesis that the EEG data were drawn from different
populations. Further, the individual Chi-squares were obtained in
approximately 3 Hz bands to determine where significant effects took
place within the 1-20 Hz population.
Seizure diary data were examined as a function of treatment
conditions to determine the possible correlation of seizure rate with
the acquisition of the dependent variables. In addition, a correlation
between seizure rate and all dependent variables was performed.
The statistical analysis was directed to determining the sig-
nificance of changes between the separate conditions of the experiment,
A-B -C-B2-(D), in the sequence of treatment shifts. The output from
the Autogen 5400 was punched onto cards and subjected to a time-series
analysis using an autocorrelation function and curve-fitting procedure.
The data in the study represented a 7 x 1761 matrix (each of the depen-
end variables times each of ten five-minute epochs foreach of the thirty-six
trials). An autocorrelation function is a correlation of a time-series
with itself. It is obtained by pairing observations of x units apart
(Gottman, McFall and Barnett, 1969). This then provided the serial
correlation of these data as a function of lag. For the baseline period
lags of one through ten were employed, while for the pre- and post-
periods lags of one were employed. The curvefitting procedure involved
fitting the data to the least squares straight lines. The difference
between slopes and between means was then calculated using as a test
of significance a series of all possible t-tests.
Changes affected by the treatment procedures were analyzed by three
procedures: 1) slopes within treatment conditions were assessed with
respect to whether each slope was negative, zero, or positive; 2) dif-
ferences between slopes within each period for each variable and,
finally, 3) the differences between the means of all baseline-
treatment combinations within each subject were calculated.
The presence of a slope significantly different from zero within a
condition indicted that a relationship existed between the treatment
conditions and the EEG variable under consideration. For example, a sig-
nificant positive slope demonstrated that a direct relationship existed
between application of the treatment and an increase in the EEG variable.
Likewise, a negative slope demonstrated that the EEG variable decreased
in magnitude over the course of the treatment condition.
The significance of the slopes was evaluated following the removal
of the autocorrelation effects. This removal of the autocorrelation
was made necessary since EEG data collected on a subject was predictive
of data produced over successive periods. Once accomplished, the data
assured the independence of repeated observations over time. Following
this, a least squares line was determined for the EEG data within each
subject, for each variable during each condition.
The autocorrelation and least squares analysis was accomplished by an
evaluation of the data with a commercially available autoregression program
(Barr, Goodnight, Soll, andHelwig, 1976). The program's output provided
ordinary least squares analysis, autocorrelation statistics and estimates
of the least squares statistics after removal of autocorrelation, to-
gether with the calculation of the significance levels of the tests.
Next, selected contrasts between slopes were calculated using a modi-
fied "t" statistic. The difference was calculated by the following equation:
"1 1 2 2 1s 1
n1 + n2 2 n1 n2
where x1 = baseline slope estimate
x2 = treatment slope estimate
n1 = number of observations in baseline condition
n2 = number of observations in treatment condition
S2 = baseline slope variance
S2 = treatment slope variance
The significance was evaluated to allow for adjustment of alpha
error as a function of multiple pairwise contrasts using t-statistics
Finally, selected contrasts between the means within each condition
were accomplished. The equation was the same as (1) above but employed
the estimated mean condition value for Xl & i2, and s & 2 terms were
replaced by the error sum of squares obtained from the autoregression
program. Again, the table provided by Games (1977) was employed to
evaluate significance levels.
The results are detailed in three major sections: 1) hypothesis 1,
frequency analysis data of pre-/post-EEG measures; 2) hypotheses 3 and
4, time-series analysis of day-by-day training data comprised of
alpha and theta changes; 3) hypothesis 2, seizure diary data and its
relation to treatment conditions.
Data with Respect to Hypothesis 1
Changes in the Frequency of EEG Occurrence <8 Hz and in the
9.02-12.05 Hz Range
This hypothesis predicted that subjects will display a frequency
shift from slow frequency ranges <8 Hz toward predominantly alpha
activity in the 9-13 Hz range. This result was predicted from the
contingencies presented in the experiments, namely, a) physiological
feedback and instructions to suppress low frequency activity in the
<8 Hz range, and b) instructions to increase the occurrence of
frequencies between 9-13 Hz (exclusive). (See Tables 2-11 for chi-
squares, number of counts and mean frequency data for all subjects.)
Subjects #4, #5, and #6 displayed a frequency shift from low
frequency activity toward faster as measured by the mean frequencies
of the pre-/post-EEGs.
Subject #1, because of a skull defect which allows high frequency
activity to be observed without normally occurring attenuation, showed
+ I I !
ML r- N- C
o r- r, 0M 0o -
C"j to LO LA r
m CD C m
C)O0 0 CM Co
r-. C O -
Ico I I I I
N N N N N
0 0 0 ---
I I I I I
e- CM 01 O> 0
-C- M Mn 0 N
LO CY C
00 C 0
LO O Q.
VI VI VI II
+ I I
M 1.0 c Co
m -K LA (n C\J
CO O 4- U N CM
00 M C
- W0 CMJ -
r-( cj o'
.CO Co C
,- 0 .0 M q O
0 MCOo 0 0
N N N N
rI I I
\j d- M- L
0 0 0
D C CD CD
S- 0V 0
n C) o L
VI VI VI II
CM 0 ) t.0 .0 CMJ
No Cnn Zz- n COo C\j
C0 _- 0 (V C\M CM *!3
0 CD C\j CM" C -
<-z 0o 0 0') 0 LO (.0
t.0 Ln rN- co rN- 1- -zr
CM1 r,- rN- -d- M C-
CO o 0) Ln CMJ o
CMi oo CM Co CD CD CD
+ + I I I
i- ^1o U n 0
00 CM 0 CM)
C) 0M 0- C- UC
i- r-.. C' cri '.
CM) r- CM
00 U- %:t
*- o0 *) *) u) oC)
N N N
I- "- 0
O CO C
VI VI 4 II
C\j t co0
m'd Co 0 k.0 in Ci
C- N COM CO C0 m
S LU- inLO
CO) Cj M0
0 Cj 0m
I I I I
r- CN 01
- C?) 1n
r- r-.n *r-
c? 00 CO Co m
i- Co u c? C
N C-- r-. e
C\J C 0 C0 a,
N N N
in CO 0
I> 0 Q.
VI VI VI II
0. L. +-
4- -t- 4- +4-
f I I + + +
o 0o c C\ CO Lm n cc
- CJ C'Oj CJ 0 0 0) CD
SWC\A( 0 0 0 (
N N N O O 0
I I I
VI VI VI II
K K 9
NC r L
0D 00 o c4
P- Co m 0) CM
LO LA CJ C Ln
u cr eL A
LO LA CMi to )
0 eoj c'. c-
4- +- +- 1-t
I I + + + + i
10 *- r-
- 0 L
O iD 0m O CD O
M C" O Ow W n -j- CM
"- 0 :- i-. CM"
0M CM -- 0 0
N N N
I I I
LO c\J C\i
0I- r CO
Cn ML CM
vI vi vi II
V Ic VI VI I
mC *C -Ic
*1n -CM '-in
i- 0'n oo '
i- LA i- 0'
C C i
- + +- +- +-
I + + + + +
3 0i 0'- C- 00 (0 0
N. N. (0 (0 (0 O
No- m -CO m m C
0 LO o oD
O0 M M C-0 00
0)0 CD -
m>0 o -
C oO '- C COC CO
C\l (n0 C L Nt M
C 0 0 0 to M r--
- U CO 0 0 0
*r- 0 O
c 0a CO
UO r- O C.
VI VI VI II
0-. Q. 4
LO ON Cr"
4- -- -i- 4-
+ + + + +
m- m O *- 0
m C) 0 LO
rN ) 0 C\- 0- U0 C)
MO C 0 0 0
I I I
V- ti u;
M Un MDOU OD O M
(U CMQ ~) r-. U) Un
U) k. Cj 'D 0 CD0 )
0 C) C ) 0
VI VI VI
C1C. I -
- 4- 4- +- +- 4-
II + + + + +
tD CM CM
CMJ 0) CMJ
M M CMJ
M r -- m
O t3 O0 O M CJ -
NJ Cr) N -- 0 0 0
II II V I
CN 4- .
LD r- O D.
L. CI V +-C
M M M 4-
CC- LO O C Cr -
hV O 0 n I--- o, r-
t-. CN i- (U
D a) CO .o r-% r) *m
00 o ao to r- 0
1n O C3- '0 r),- 0,
* .)o >
- 4- 4- 4- +-
SI + + + + +
C I- <- C\j LN CO CO C
3 n 0 oo Ci CO em
0 LO O ) LA m
o C in O> in M
C CM M CD C
LA CM LA
CO O O \ M r-.
M MN CD
-- CCM CCO- CO CMJ
-- n C' r-0 0 0
N N N
3 O -- (.
o o ,-
NA 0 )
II II V I
C\J 4- 0 .
O D oC
Sc Vo >,
VI VI VI II
0. 0. +-
r- CO 00
O 0 0 O0
o r-. 1- LA
a significant reduction in average frequency. For this subject this
shift was considered a normalizing one. He, unlike the other subjects
on the initial EEG, had not manifested overall slowing but instead
produced an increased amount of faster activity in the 7-13 Hz range.
Subject #2 displayed a slowing of activity with pre-/post-
measurement, and this result does not appear to support the hypothesis.
Dividing the overall frequency pattern into approximately three Hertz
bands allowed a finer discrimination to see where shifts did or did
not take place.
Subjects #4, #5, and #6 displayed a significant diminution of low
frequency responses in the frequency bands 1.0-3.01 Hz and 3.01-6.02 Hz
Subjects #1 and #2 did not exhibit the predicted change in these bands,
instead their post measures showed an augmentation of slow frequency
responses. Both subjects #1 and #2 did, however, evidence a signifi-
cant diminution of response in the 6.02-9.10 Hz range.
Subject #4 was the only person who displayed a significant in-
crease in the number of counts in the 9.10-12.20 Hz range. Subjects
#5 and #6 produced a positive but nonsignificant count increase in
this band. Subjects #1 and #2 exhibited a significant decrease in
the number of counts, contrary to hypothesis 1.
Other Associated Changes in EEG Pattern Suggested by Frequency
Though not reinforced, a significant augmentation in the number
of counts was demonstrated within the 12.2-15.2 Hz and 18.2-20 Hz range
for subject #5 on both channels. Subjects #4 and #6 demonstrated a
significant increase in the 12.20-15.15Hz range in channel 1. Though
nonsignificant, subjects #4, #5, and #6 produced a consistent trend
toward the faster frequencies from 9.10 Hz on. Subjects #1 and #2
demonstrated the opposite effect; subject #1 evidenced a significant
diminution on channel 2 in the ranges of 9.10-18.18 Hzand on channel 1
in the 12.20-20.0 Hz range. Subject #2 displayed a significant slowing
in the 9.10-12.20 Hz range and a nonsignificant slowing trend in the
12.20-18.18 Hz range.
Since the Chi-square changes represented both distributional and
level changes, the mean frequency shift was more instructive in
describing those changes. It did, however, reflect the combined
effects of all frequency bands and, in the specific case of alpha
diminution, may have distorted the picture.
Unfortunately, in dealing with dependent data, percentages were
employed and a shift in one frequency band was expressed as a change
in another. If the EEG is perceived on a continuum from hyper-
synchrony (many synaptic potentials occurring together) to synchrony
(with synaptic potentials largely out of phase), attempts to shift
toward a midfrequency range may result in diminished desynchronization.
Data with Respect to Hypothesis 2
Hypothesis 2 made the following predictions: 1) overall, seizure
incidence will diminish over time and be significantly less at the
completion of the study, and 2) this decrease will be a function of
the training condition (see Table 12 and Appendix B) and the abundance
of alpha and theta activity throughout all conditions (see Table 13).
Overall Seizure Incidence
Table 12 illustrates the test of the hypothesis that no shift
in the seizure rate occurs throughout the study. That is, the null
CO C CUJ Ll 0
co r-- :I- r"
< u Cn L.) Co Q
1.0 L- CM'
d- d- CYCM
MO Ln CMJ < r
co Co Cr> ir CMj
O0 O CO0 m
cM. f r- o
O* 'D .- C\j o
( o 0cn or x-o
Ln C I- r-
- *- r- 00o o
n *t- 0
CMir- tjI LA
-.r-. oo I:a-
N M N
SC *r- C 3
C w) 4--) a) L)
*- E c E
i4-) 04-) (0
-- M U (0 4-
V) a) C w} a)
r( I- 0 S- .C
0 -- Z 1-I--
4-) C 4-) M-
)C *r- C a
a) C O .) 4-
I a- C 0) a)
to S.- 0 S- r-
CC0 = 1- -
1. 1'0 .- L A
LA LO CMj -
-o m o\ 'co
mD C M- L- Co LA
LO '.O- X)
4-' C 4-)
a C *- C
C Ul 4- a)
*' E C E
( 01) C _a)
to 0 I-
Co 1- 21- l'-
co C -
) *r- :
> WU ,
SrMo -I o
-0 o ",-.
.rl- (C) kD m-c
-t 00C o 0m
MOn o 00 CM
i n I b Z1
- CM. Oo CM)
OO 00 Cj
=- -- N CJ
- -C C J0CoI
3- O. -rO0o-) CM
C4- 0 4-)
a) (U U rt 4-'
n Q) rc QU) W
t01 .*- c0
* E c--
0) 0 10 -l
10s-o s ^
00 CM ir-
4-) C 4-)
U C *r- C
C C0 )4-' (U
*- E C E
4-- 0 4-
C13 I-- = I--
LO LO L
Ln LO *at
0 0 LA
o o o
C, ai -W
0 U- u
0 > 0 0
> -c .,c
ct 5- i-
VI VI VI VI
Q. DQ a- 0.
4 4 K -X
hypothesis that the treatment administered in a condition had no
effect upon seizure rate. The expected value reported for each condi-
tion in the table then is the product of the total number of seizures
each subject had throughout the study multiplied by the percentage
of total time occupied by each condition. This test appears con-
servative since the influence of the experimental conditions was of
interest rather than a pre-post test of seizure rate. Appendix B
presents a graphic summary of the observed and expected Chi-square
values from Table 12 for each subject.
Subject #1 produced a nonsignificant Chi-square indicating no
change in seizure incidence over the course of treatment. Subjects
#2, #4, #5, and #6 all displayed significant Chi-square values indicat-
ing a change in the frequency of seizures over the time course of
training. Taking into account the average seizures per day, subjects
#2, #4, #5, and #6 displayed an overall decrease. Subject #1, in
effect, displayed no change in the average seizures per day.
Seizure Incidence as a Function of Training Conditions
During baseline condition A, subjects #4, #5, and #6 produced a
significant departure from expectation by exhibiting a greater number
of reported seizures as measured by the average seizure per day figures.
This result was in keeping with the predictions of hypothesis 2 since
the baseline period (A) was assumed to measure the frequency of seizures
prior to intervention. Subjects #1 and #2 evidenced nonsignificant
results during this condition, which indicated no elevation in seizure
activity during the baseline period.
During the first treatment condition (B1), subject #4 demonstrated
a highly significant departure from expectation in seizure frequency.
This result was consistent with the predictions of hypothesis 2. Sub-
jects #2, #5, and #6 produced a nonsignificant diminution in seizure
incidence during treatment B1. Subject #1 displayed an increase in
seizure production during this condition, a result which was con-
tradictory to the expectations of hypothesis 2.
During condition C, the noncontingent feedback phase, subjects #2,
#4, and #6 displayed significant decreases in seizure rate--a result
not predicted during this condition. Subjects #1 and #5 displayed
nonsignificant decreases in seizure frequency--a result that was also
not predicted by the noncontingent feedback of condition C. This
matter will be explored further in the next chapter.
During the second feedback condition (B2) only subjects #4 and
#6 evidenced a significant negative correlation. For these subjects
these results indicate a decrease in seizure frequency.
Of the three subjects taking part in condition D, the direct sup-
pression of theta activity without concurrent alpha enhancement, subject
#6 was the only one to display a significant diminution in seizure
incidence. Subject #1 produced a significant increase in seizure
frequency. From the standpoint of hypothesis 2, only subject #6's
result was predicted.
Table 13 presents the correlations of EEG variables with seizure
frequency over all subjects and each individual subject. For all sub-
jects there was a significant negative correlation between alpha per-
centage and seizure incidence. As alpha percent diminished, the
seizure rate increased. On an individual basis, subjects #4 and #6
evidenced significant negative correlations while subjects #1, #2, and
#5 produced negative but not significant ones.
For all subjects there was a significant positive correlation be-
tween average frequency and seizure incidence. In effect, as the aver-
age frequency increased, the rate of seizures also increased. On an
individual basis this result was particularly displayed by subject #4.
Though the overall correlation between average amplitude and
seizure frequency was not significant, subjects #2, #4, and #6 all
displayed significant positive correlations. This indicated that as
the average amplitude in their EEG waves increased there was a
similar increase in seizure frequency.
In reference to theta percentage and theta count subject #2
displayed a significant positive correlation indicating that there
was an increase in seizure rate as theta production increased. This
was the only subject for which this was true.
Data with Respect to Hypotheses 3 and 4
Hypotheses 3 and 4 were evaluated in two ways: one, the
significance of the slopes within each condition was determined.
These data allowed statements to be made as to the rate of change
within each condition testing it against the null hypothesis that
slope = 0. Second, the significance of the difference in slopes
between conditions and the difference in means was determined. These
data, too, allowed statements to be made as to the possible differences
in treatment versus baseline conditions testing against the null
hypothesis that the differences between slopes and mean = 0 (Kazdin,
1976). The slope data obtained within treatment conditions was
presented first, followed by the combined presentation of the contrasts
between the slopes and means for baseline versus all treatment
Data with Respect to Baseline Condition A
Condition A stated that during the initial three-trial baseline,
subjects will display a zero slope (see Appendix C within condition
for data). Subjects #1 and #5 produced the predicted nonsignificant
slopes indicating no change in alpha percent during the baseline
period. Subjects #2, #4, and #6 displayed significant slopes with
subject #2 producing a positive one, which indicated an increase in
alpha and subjects #4 and #6 producing negative slopes which indicated
a diminution in alpha percent.
Data with Respect to Condition B1
During the eighteen-trial treatment condition, subjects will
display a positive slope and an increase in level change from baseline
condition A. In this context, level and mean represented the same
concept--namely, that a change occurs in the mean or average level of
a given dependent variable between one condition and other. It should
be noted that this definition differs from the one usually applied in
the classical time-series literature, which refers to a level change
as occurring when a discontinuity exists between the data at the
point when a treatment is introduced (Hersen and Barlow, 1976).
This condition consisted of a five-minute segment pre-/post-
alpha feedback where no feedback was forthcoming. These segments
were termed pretreatment B1 and posttreatment Bl, respectively.
Between these no-feedback segments there was a forty-minute alpha
enhancement condition to be called treatment B1.
During the pretreatment B1 period for within condition slopes,
subject #5 displayed no change in the rate of alpha production, a
result that was not coincident with a combined feedback/no-feedback
condition as one might have expected, as one might have expected a
gradual increase in alpha acquisition over time. Subjects #1, #2,
#4, and #6 all displayed a significant increase of alpha percent, a
positive slope. This result was coincident with the hypothesis.
During the treatment B1 condition, subjects #1, #4, #5, and #6
displayed the predicted enhancement of alpha percent during this
feedback phase. Subject #2 yielded a nonsignificant slope indicating
no change in alpha percent during this phase.
During the posttreatment B1 condition, subjects #1, #2, and #6
displayed nonsignificant slopes indicating no change in alpha produc-
tion. As with the expectations of pretreatment B1, this result was
not in keeping with the hypothesis. Subjects #4 and #5 produced a
significant positive slope indicating an increase in alpha percent.
Because these subjects also produced a significant treatment B1 effect,
this carryover to an immediate posttreatment no-feedback phase may
well have been a function of learning and, therefore, consistent with
the predictions of hypothesis 3.
Data with Respect to Condition C
During the four-trial noncontingent condition, subjects will
produce a negative slope as compared with Condition B. As with the
treatment B1 condition, the noncontingent phase was divided into
three parts: 1) pre-noncontingent phase of five minutes with no
feedback; 2) forty-minute noncontingent phase, when sham feedback was
given; and 3) post-noncontingent phase of five minutes when no
feedback was forthcoming.
Subject #1 yielded a significant negative slope, indicating a
diminution in alpha percentage during this phase. Given this sub-
ject's prior history of a skull defect and inability to filter out
faster waves, this result might have been expected. During this time
this subject began again to produce fast activity very similar to that
produced during condition A. Subjects #2, #4, #5, and #6 yielded
During the noncontingent feedback phase, subjects #1 and #5 pro-
duced significant negative slopes indicating a diminution in alpha
production as a result of this false feedback stage. This result was
in keeping with the tenets of hypothesis 3. Subjects #2 and #4 yielded
nonsignificant slopes, indicating no change in alpha production as
compared with baseline measures. Subject #6 displayed a significant
positive slope which indicated an increase in alpha percent.
During the post-noncontingent phase, subjects #1, #2, #5, and #6
all produced nonsignificant slopes indicating no change in alpha wave
percentage. These results were predicted especially in the case of
subject #5, who produced a significant diminution in alpha in the
prior phase and a nonsignificant slope in the pre-noncontingent phase.
These results, taken together, indicated that for this subject the use
of noncontingent feedback had a direct effect upon his response. When
he was released from this feedback, he reverted to a no change status.
Subject #4 produced a significant negative slope indicating a
continued diminution of alpha percent during this latter phase.
Data with Respect to Condition B2
During the six-trial contingent alpha enhancement condition
subjects will display a positive slope and a significant increase in
level from baseline condition A.
During the pretreatment phase subjects #1, #2, #4, #5, and #6 all
yielded nonsignificant slopes indicating no change in alpha production.
During the contingent feedback phase, subjects #1, #2, and #4
yielded nonsignificant slopes indicating no change in alpha production
as a result of feedback. Subjects #5 and #6 produced significant
negative slopes demonstrating a decrease in alpha wave production.
These results are contrary to the direction predicted by the contingent
phase of hypothesis 3.
During the posttreatment phase all subjects yielded nonsignificant
slopes, indicating no change in alpha production during this phase.
Data with Respect to Hypothesis 4
Data with Respect to Condition A
During the initial three-trial baseline, subjects' data level will
be stable and a least-squares line fitted to the data will have a
Subjects #2, #4, and #6 displayed the predicted nonsignificant
slopes indicating no change in theta percent during the baseline
period. Subject #1 produced a significant positive slope indicating
an increasing amount of theta production in the absence of feedback.
Subject #5 exhibited a significant negative slope indicating a
decreased theta production during the baseline period.
Data with Respect to Condition B1
During the eighteen-trial treatment condition, subjects will
display a negative slope and a decrease in level from baseline
This was a tripartate condition consisting of a five-minute
segment pre-/posttheta suppression where no feedback was forthcoming.
These segments will be termed pretreatment B1 and posttreatment B1,
respectively. Between these no-feedback segments there was a forty-
minute theta suppression condition to be called treatment B1.
During the pretreatment B1 phase subjects #1, #2, #4, #5, and #6
all displayed nonsignificant slopes. This result was not in keeping
with hypothesis 4. Because of the contiguity in time of treatment Bl,
a gradual increase in slope indicating a learned generalization was to
have been expected.
During the treatment B phase, subjects #1 and #4 displayed a sig-
nificant diminution in theta production, a result which is coincident
with hypothesis 4. Subjects #2, #5, and #6 evidenced nonsignificant
slopes indicating no change in the amount of theta produced.
During posttreatment B1, all subjects displayed nonsignificant
slopes indicating no change in theta production during this no-feedback
phase, again inconsistent with the hypothesis.
Data with Respect to Condition C
During this four-trial noncontingent phase, subjects will produce
a positive slope and a decrease in level from baseline condition A.
Like the treatment B1 condition, the noncontingent phase was
divided into three parts: 1) pre-noncontingent phase of five minutes
with no feedback; 2) forty-minute noncontingent phase where feedback
was supplied; and 3) post-noncontingent phase of five minutes with no
During the pre-noncontingent phase all subjects produced non-
significant slopes indicating no change in theta production, a result
coincident with hypothesis 4.
During the noncontingent feedback phase, subject #5 displayed a
significant negative slope indicating a diminution in the amount of
theta produced. Subjects #1, #2, #4, and #6 evidenced nonsignificant
slopes indicating no change due to noncontingent feedback.
During the post-noncontingent phase, all subjects displayed
nonsignificant slopes indicating no change in the production of theta.
Data with Respect to Condition B2
During the six-trial contingent phase, subjects will display a
negative slope and a significant decrease in level from baseline
As in the B1 condition, B2 was contingent and divided into three
phases: 1) a five-minute pretreatment phase with no feedback;
2) forty-minute treatment phase with theta suppression; and 3) a
five-minute posttreatment phase with no feedback.
During the pretreatment B2 phase, subject #1 produced a significant
diminution in theta percent, and subject #5 displayed a positive slope
which indicated an increase in theta production. Subjects #2, #5,
and #6 produced nonsignificant slopes which indicated no change in the
production of theta.
During the treatment phase of B2, subjects #1 and #5 displayed
significant negative slopes indicating a diminution in theta production
during this phase. This result was consistent with the tenets of
hypothesis 4. Subjects #2, #4, and #6 displayed nonsignificant slopes
indicating no change in theta production due to the combined alpha
augmentation and theta suppression.
During the posttreatment phase of condition B2, subjects #2 and
#5 produced negative slopes indicating a decrease in theta production
in this phase. Subjects #1, #4, and #6 displayed nonsignificant slopes
showing no change in theta production.
Data with Respect to Condition D
During the three- to four-trial theta suppression condition,
subjects will display a negative slope and a significant decrease in
level from condition A.
During the pretreatment phase of condition D, subjects #1, #2,
and #6, the only ones taking part in this condition, all displayed
nonsignificant slopes indicating no change during this phase.
During the treatment phase of condition D, subjects #1 and #6
both evidenced a significant negative slope indicating a diminution
in the production of theta. This result was consistent with
hypothesis 4. Subject #2 produced a nonsignificant slope showing
no suppression in theta production.
During the posttreatment phase of condition D, all subjects
displayed nonsignificant slopes indicating no change in theta during
Between Slope and Mean Differences
Though all pair-wise comparisons were determined, only the dif-
ferences between conditions A-B1, A-C, A-B2 and A-D, that is, the
differences in baseline and treatment conditions, were reported for
alpha percent and theta percent. Appendix D presented the selected
pair-wise contrasts between these conditions for slopes and levels.
Data with Respect to A-B1 Alpha/Theta Comparisons
The pretreatment phase condition B1 yielded a significant ac-
celerated alpha slope for subjects #1, #4, and #6. Subject #2 dis-
played a deceleration in alpha production in this comparison.
During the treatment phase subjects #1, #4, #5, and #6 evidenced
a significant accelerated alpha response slope as compared to baseline.
Subject #2 yielded a significant deceleration in slope in this
During the posttreatment B1 phase subjects #1, #4, and #5
yielded slopes that were accelerating significantly when compared to
In reference to the slope of theta production during the pre-
treatment B1 phase, subject #1 produced a significant decelerated slope
which indicated that this phase had an effect on changing responding
in a negative direction. Subjects #2 and #5 displayed an A-B1 slope
comparison which indicated a greater acceleration of theta as compared
During the treatment phase the slope of theta production was
significantly decelerated for subjects #1, #4, and #6. Subjects #2
and #5 displayed significantly accelerating slopes which indicated
an increase in theta production as compared to baseline.
The posttreatment A-B1 comparison yielded a significant decelera-
tion in slope for subject #6. Subject #5 produced an acceler-
ated slope which indicated an increase in theta responding.
Data with Respect to A-C Alpha/Theta Comparisons
The pre-noncontingent phase yielded a significant deceleration in
slope for alpha percent for subjects #1 and #2. Subjects #4 and #6
displayed an acceleration in alpha production as compared to baseline.
The noncontingent phase of condition C yielded a significant
deceleration in the slope of alpha percent for subjects #1 and #2 as
compared to baseline. Subject #6 displayed significantly
accelerating slopes which indicated a continued increase in alpha
The post-noncontingent phase yielded a significant deceleration
in the slope of alpha for subject #2. Subject #6 displayed an
accelerated slope as compared to baseline.
The A-C slope comparisons of theta production during pre-
noncontingent phase of condition C yielded a significant deceleration
for subject #1 as compared to baseline. Subject #6 displayed an
accelerated slope which indicated greater theta responding as compared
to baseline. Subject #5 displayed no significance difference in
slopes but did evidence a significant level difference, indicating
greater theta responding though no change in direction of response.
During the noncontingent feedback phase subject #1 displayed a
significant deceleration in slope which indicated a decrease in theta
productivity as compared to baseline. Subject #5 displayed a lessened
deceleration as compared to baseline. Subject #2 produced an acceler-
ated comparison which indicated an increase in theta responding.
During the post-noncontingent phase subject #1 displayed a
difference in slopes that indicated a deceleration of theta produc-
tion. Subjects #5 and #6 produced a significant acceleration in
theta production as compared to baseline.
Data with Respect to A-B2 Alpha/Theta Comparisons
The pretreatment phase of condition B2 yielded a significant
acceleration in the slope of alpha production for subjects #4, #5,
and #6. Subject #2 yielded a slope comparison which indicated a
deceleration in alpha responding as compared to baseline.
During the treatment phase of condition B2 subjects #4, #5, and
#6 yielded significant accelerating slopes as compared to baseline.
Subject #2 displayed a deceleration which indicated a lessening of
alpha production in this comparison.
The posttreatment B2 comparison yielded a significant accelera-
tion in slope for subjects #4 and #6. Subjects #2 and #5 displayed a
comparative deceleration in alpha production.
In reference to the A-B2 slope comparisons of theta production
during the pretreatment phase, subject #1 displayed a significant
deceleration in slope as compared to baseline. Subjects #2, #4,
and #5 produced A-B2 slope comparisons that indicated an acceleration
in theta production in the B2 pretreatment phase.
During the treatment phase subjects #1, #4, #5, and #6 displayed
significant decelerations in slope indicating a lessened production of
theta. Subject #2 displayed a significant slope difference
which indicated an acceleration in theta percent in the comparison.
The posttreatment A-B2 comparison yielded a significant decelera-
tion in responding for subjects #1 and #6. Subject #5 displayed
a difference in slopes that indicated an acceleration in responding
during the posttreatment B2 condition.
Data with Respect to A-D Alpha/Theta Comparisons
Since only three subjects (#1, #2, and #6) took part in the D theta
suppression condition, A-D comparisons were confined to them.
Subjects #1 and #2 displayed a significant deceleration in the
slope of alpha percent during the pretreatment phase. Subject #6
yielded a slope comparison which indicated a significant acceleration
in this phase as compared to baseline.
During the treatment phase subjects #1 and #2 again yielded a
significant deceleration in alpha production as compared to baseline.
Subject #6 again yielded a significant acceleration in alpha
The posttreatment phase yielded a significant deceleration in
the slope of alpha percent for subjects #1 and #2. Subject #6 displayed
an acceleration of slope which indicated an increase in alpha responding
as compared to baseline.
During the pretreatment phase of condition D the A-D comparison
in slopes for theta production yielded a significant deceleration in
slope for D condition as compared to baseline for subject #1.
Subject #2 yielded a t-ratio of slope comparisons that indicated that
there was an acceleration of theta production.
During the treatment phase the A-D slope comparison yielded
significant deceleration in responding for subjects #1, #2, and #6
which indicated a lessened production of theta during this phase.
The posttreatment phase of the A-D slope comparisons yielded a
significant t-ratio which indicated a significant deceleration in the
slope of theta for subject #1. The remaining two subjects yielded
Of the fifty-four individual pair-wise contrasts for alpha
percent, none of the mean differences between slopes were significant.
This lack of difference, combined with the fact that the vast majority
of the differences between slopes were significant, harkened to a
change in direction of the individual slopes but not to level. This
result may well have been due to a baseline period that did not allow
differences in baseline versus treatment conditions to be discerned.
Of the fifty-four individual pair-wise contrasts for theta per-
cent,only three (subject #5 A-C, significant increase in slope and
level; subject #5 A-B2, significant increase in slope and level; and
subject #2 A-B2 significant increase in theta percent) proved to be
significant. As with measurement of alpha percent, the inadequacy
of baseline data may well have accounted for this result.
If physiological feedback can be considered a slowly learned
discrimination task, the results of significant slopes and nonsignifi-
can level might have been expected. A skill that required time to
develop would have produced stable maximal levels before the cumulative
effect of training produced significant differences in the means of
Conditions. The fact that there were many significant within condi-
tion slope changes was seen as supporting this contention.
The present study provides additional evidence for the efficacy
of EEG physiological feedback training in the midfrequency range as a
means of controlling seizures. To explore this matter further, this
chapter will be divided into three sections: first, a discussion of
individual hypotheses and the basis for their support or rejection;
second, critique of the experimental design and suggestions for
improvement; lastly, suggestions for future research.
Discrepancies Between Hypotheses and Results
Hypothesis #1 predicted an increase in EEG frequency during
feedback training. The overall results support this contention in
three of five subjects (#4, #5, #6). These subjects increased their
average frequencies in this pre-/postmeasurement condition.
Subject #1 displayed a decrease in average EEG frequency across
both bipolar electrode sites (T3-T4, C3-C4) in the pre-/postfrequency
analysis. As stated before, this subject's skull defect caused him
not to attenuate higher frequency activity, and, hence, the pretraining
EEG displayed this overabundance of fast-wave activity. The alpha
band (8-13 Hz) is essentially the middle range of the typical frequen-
cies recorded in the human EEG; an individual with predominantly
desynchronized beta activity when presented with the contingencies
of alpha feedback is likely to decrease his frequency to maximize
feedback in accordance with established set. The question that
is of course left unanswered in this subject's case is whether
diminishing the frequency reduces the subject's seizure rate. Since
no significant decrease in seizure rate occurred, it appears his
response to the task requirements was not beneficial.
Subject #2 was unable to increase mean frequency pre/post.
The conclusion one draws from this and the fact that he was also
virtually unable to control alpha and theta training contingencies is
that the procedure, as it is presently constructed, does not work for
him. This subject had a successful diminution of seizure rate only in
the C or noncontingent condition. Heightened arousal level resulting
from stimulus novelty (orienting reflex) cannot account for this
result in the subject's response because his average frequency
decreased during noncontingent periods.
Seizure Incidence and Treatment Condition Analysis
The finding that seizure rate was positively correlated with
average frequency is a potentially important result. This is in
contrast to Sterman's most recent statement (1978) where he indicates
an increase in average frequency is negatively correlated with seizure
rate. In discussing this discrepancy it is most important to note
the differences in the subject's seizure types chosen for the two
studies. Sterman chose subjects heterogeneous in seizure type ranging
from the centrencephalic to focal type. The present study chose
subjects only with focal seizures because it was thought best to
test the role of physiology feedback on a homogeneous population. It
is clear that the mechanism of seizure propagation is quite different
in centrencephalic versus focal seizures (Jasper, 1969). To attempt
to make any generalizations as to the efficacy of any one approach,
a distinctly homogeneous population is a prerequisite.
From the standpoint of focal seizure patients, the results indi-
cated an increase in seizure rate with an increase in average frequency.
As the variability of alpha production decreased, the seizure rate also
increased. From the standpoint of the malleability of the control
mechanisms of the brain, this result supports the notion that a
lessening of synchronous activity reduces the ability to inhibit
seizures for these subjects (Jasper, 1969). In effect, as the
thalamocortical synchronization reflected in alpha percentage decreases
during the production of faster, essentially desynchronous activity,
seizure rate increases. This result argues for a midrange hypothesis,
in effect a normalization, for individuals with focal motor seizures.
Alpha percentage was significantly and positively correlated with
feedback in the B1 treatment condition for four out of five subjects.
These results indicated an ability to acquire the alpha response as a
function of the treatment condition. When noncontingent reinforcement
was instituted, there was a significant within condition shift and a
between condition slope difference for three of these four subjects.
This same overall result was not forthcoming for the second contingent
condition (B2) and argues to the brevity of this condition as a cause
for its failure. This matter will be explored further in the section
on design critique.
Subject #2 displayed no effect from alpha acquisition training
during any of the treatment conditions. There were no readily explain-
able reasons for this result; this subject assiduously maintained
appointments, understood the task requirements and very much wanted
improvement. As will be stated later, the anomalous nature of this
subject's results argues toward the institution of an individually
designed treatment program using a prolonged baseline as a means of
discerning the appropriate contingencies for reinforcement.
Two of the five subjects were able to suppress theta occurrence
during the treatment B1 condition. The instability of this suppression
is mirrored by the fact that neither of them had any transfer to
either the pretreatment B1 phase or the posttreatment B1 phase. It is
well to view the general lack of success in the suppression of theta in
view of the pre-/posttraining EEGs. Three of the four subjects who
were successful in seizure abatement produced significant decreases in
the 1.0-3.01 and 3.01-6.02 Hz band (#4, #5, #6). None of these subjects
produced a significant change in the 6.02-9.10 Hz range. It is sug-
gested that the abnormally large number of counts in the 1.0-6.02 Hz
range to begin with as compared to posttraining constitutes an important
normalizing of these subjects' EEG patterns. In effect, these subjects
began with slower average frequencies than was anticipated by the
fourth hypothesis. Their training lessened the occurrence of delta
activity. It is likely that more extensive training would have had
an eventual effect upon the higher theta range (6-9 Hz).
Design Critique and Suggestions for Improvement
In discussing the strengths and weaknesses of the design we must
begin by realizing that the results taken as whole indicate an
acquisition of alpha without a similar diminution of theta.
From the standpoint of design the major weakness of this study
lies in an insufficient extended baseline period of EEG recording. If
these periods were sufficiently extended, one would expect a non-
significant slope indicating a consistent response mode from the
subject. Two of the five subjects displayed significantly negative
slopes in alpha percent and had this trend been allowed to continue
may well have enabled the significance of the mean difference between
baseline and contingent feedback (Bl) to be greater.
This statement is particularly important in regards to the mean
differences between conditions. Given that the vast majority of the
t-ratios for the differences between baseline and treatment conditions
were significant, a stable baseline period may well have enhanced
the possibility of significant differences in level,too.
One method of evaluating these changes from the baseline would be
to employ a multiple baseline procedure. Essentially this entails a
number of responses identified and measured over time to provide
baselines against which changes can be evaluated. When the baselines
are established, the experimenter then applies a treatment condition
to one of the behaviors and observes little or no change in the other
variable (Baer, Wolf and Risley, 1968). In this way significant
results would be more readily determined through the contrast of
concurrent baseline and treatment conditions.
Next, there appeared to be significant carryover effects with
measures of the dependent variables. In some cases, seizure rate
continued to remain depressed during the noncontingent conditions as
it had during the contingent ones. There are two aspects of the
carryover effect that must be explored: one, its relationship to
learning, a positive aspect and, second, its duration indicating the
need for a longer noncontingent period.
First, if the subject is being trained to respond under stimulus
control, one would expect some form of generalization outside of the
training situation. If the subject can only inhibit seizure production
during training trials, this procedure would be of little utility in
his everyday life. This carryover then can be considered a partially
positive result because of the presumed learned nature of feedback
Second, in attempting to demonstrate an effect one would assume
that by preventing the connection between EEG production and the
physiological feedback signal responding would return to baseline
levels. This did not take place presumably because tne noncontingent
condition was of insufficient duration to reestablish the baseline
levels. Yet, the only method of determining the permanence of the
changed level would be to utilize a prolonged noncontingent feedback
Lastly, in dealing with noncontingent manipulations in seizure
reduction, the ethical dilemma arises of reintroducing seizure activity
where it may have attenuated. In dealing with human subjects who
experience the debilitating effects of seizures, one sees a clash
between the very real necessities of scientific inquiry and the best
interests of the subject.
Why was alpha activity significantly controlled and not theta?
From the standpoint of task complexity both alpha acquisition and
theta suppression can be confidently considered fairly simple dis-
crimination tasks involving markedly different stimuli. It would seem
that it is easier to enhance an ongoing response set than to inhibit
it. In this dual task we are looking at the acquisition and recogni-
tion of a stimulus and the subject's response. These are two very
different tasks; the recognition of a stimulus may be far easier than
actually responding appropriately. Response differentiation could
have been facilitated by reinforcing each response individually over
time and then combining feedback signals.
Suggestions for Future Research
The aim of the present study was two-fold: one, to evaluate the
proposition that feedback of midrange EEG frequencies of the EEG via
synchronous alpha wave enhancement and theta wave suppression was an
efficacious means of seizure control; two, to attempt to develop a
procedure that was clinically viable in the sense of being able to
diminish the time needed to train an individual to attenuate seizure
In any suggestions for future research the aforementioned
methodological considerations are prominent and provide many of the
suggestions for part one of this section.
In devising future studies, one method of deciding upon the most
efficacious dependent variables is by correlating seizure diary
information rate with EEG parameters over a prolonged baseline. When
this is done, one would select only those EEG variables for reinforce-
ment which are positively related with seizure rate. This might well
permit the optimal choice of EEG variables for an individual subject
which, when reinforced or suppressed, would lead to seizure diminution.
This procedure would lend itself to an empirical evaluation of EEG
parameters, which when modified would lead to normalizing the EEG
Certainly an important covariant in a study of this nature is the
subject's present and prior history of adherence to an anticonvulsant
regimen. In this regard, it is suggested that along with a prolonged
baseline of EEG and seizure diary data anticonvulsant blood levels be
included prior to and throughout the course of training. Richens
(1976) has indicated that the very fact of telling subjects that you
are able to monitor the level of medication changes their rate of
compliance in a positive direction. If, during the course of physio-
logical feedback training, it is seen that seizures are beginning to
attenuate, it would be well to monitor these levels closely to see if
there has been any change due to increased compliance. In effect,
what is argued for is a multidisciplinary approach that combines the
medical approach of diagnosis and anticonvulsant therapy with that of
the behavioral, which seeks to modify the ability of the individual
to become aware of and control pertinent physiological states that
are antithetical to seizure production.
In developing a consistent, long lasting program for seizure
control, the role of home feedback equipment is an important adjunct.
Sterman (1978) utilized such devices and 4-channel strip chart
recorders which kept track of the time of each session, its duration,
and the number of rewards obtained. As the advances of technology
make this a more cost effective approach, the benefit to the subject
and investigator is great. The subject has a means of training
available to him that increases the ease and possibility of compliance.
For the investigator data are collected in the subject's natural en-
vironment, and the problem of generalization from laboratory to home
is considerably lessened.
Another important subject covariant is that of expectancy of
success. Each of the subjects had experienced a prolonged and chronic
illness which modified their life in significant ways. Stroebel and
Glueck (1973), in discussing the expectancy or placebo effects common
to all therapies, contends that individuals who acquire and maintain
a realistic set of expectancies (basically, one that views the
procedure as possibly beneficial within the constraints of what is
possible) will have a greater chance of success than those who either
over- or undervalue the possible outcome of the procedure. Stroebel
and Glueck (1973) consider simple, well presented information as the
backbone of a realistic set of expectancies. It is incumbent upon
any research in this area to devise a means of gauging the subject's
initial knowledge and attitudes about the disorder and the training
procedure and, then, to attempt to increase realistic expectancy by
a systematic program to educate and inform him in a step-by-step
fashion. During this procedure it is important to have an ongoing
method to measure expectancy so as to gauge the effectiveness of the
TRAINING TAPE SCRIPT
As examples of what you may expect to hear, short excerpts, previ-
ously recorded, will be played for you.
First, a tone stays on 20 percent of the time. You will notice
the tone varies both in loudness and pitch and goes on and off
Next, a tone with 50 percent success. Notice also that the tone
varies in loudness and pitch and stays on about 50 percent of the
Now, a tone that stays on 100 percent of the time. Here the
tone stays on nearly all of the time but slightly varies in terms of
pitch and loudness.
If you become sleepy or produce slow waves associated with
seizures, an alarm will sound.
This alarm will remain on as long as you are producing these slow
waves. Your job is to keep the alarm off and the sound on.
If you move around very much, a crackling sound called artifact
will be heard.
Try to remain as still as. possible so as not to produce this dound.
Again, try not to become sleepy and work at keeping the sound
on and the alarm off. This session will begin now.
CHI SQUARE PLOT OF OBSERVED VS. EXPECTED SEIZURE FREQUENCIES
Chi Square Plot of Observed
vs. Expected Seizure Frequencies
for Subject #1
195 expected -
CoNoToS: A B1 C B2 D
Chi Square Plot of Observed vs. Expected Seizure Frequencies
by Condition for Subject #2
by Condition for Subject #2
Plot of Observed vs. Expected Seizure Frequencies
by Condition for Subject #4
expected ,-- I
23D- observe m
tI expected --
..CONDITIONS,. A B1 C B2 D
Chi Square Plot of Observed vs. Expected Seizure Frequencies
by Condition for Subject #5
by Condition for Subject #5
Chi Square Plot of Observed vs. Expected Seizure Frequencies
by Condition for Subject #6
CNoDII.ONSo A B1 C B2 D
Chi Square Plot of Observed vs. Expected Seizure Frequencies
by Condition for Subject #6
SUMMARY TABLE OF SIGNIFICANT SLOPES
- ( V) ( ( (A) (AI (
k0 < V) Le) V) -Z- Ln
00 2 10100Z:- =
COa 2Z Z Z.- Z
o c'j 10L
n- OOVV Ln n
S(OJ CO OV :: -/(
zi> Cr) C z r
C- CM I A A
o r-> Cn o(A (A On co
10CM C0 2 :2o0
U0 0 00 L
*X 0 X
*X *X *X *X
10 Z Z:
V) r-I V
S I C
O OO I
U >", Q---
:r U -0 F- 0
() 4-) I-
4-) :3 -- 4- --
SC -- C
0) ). CL UJ 4-)
U S- EU CL
S- U- c< S- =3 =o ce-
-C S- S- 4J 4-) LL
,-> > J= -C
Scc 0 CC
VI vi VI VI
L (CL C .
?) C?) L) ) (Ni
cV) c( c ) cV c) c)
= = = = = z2 =
c- m-- w 0
C LLQ C
C) C C) C) C) C) ) V)
co o o o c c c
cr 0S Q: 0 w. c<
0- L- cC Q. )
<< cC I- < < >
O 0 O
o co co -0 c 0 co
I I I I I
100- Cl (n CO )
000 Z Z 0
C LO. 03 0 )
0 C C C C CO C
0Q Lj Q. L(-
o) t 1.0 L) o) I) V)
) V V) )) L) ) z )
L) LI) L) L) ) L) V)
V0 >o V V)V)
z z z z z z zn
ooU oooo oo
McK C>< >
LI) L) I) ) LI) L)
s- L I clL
CI" 0 cc 0 0
I, c ) LIr (n ) L)C
-0 'O -IC
) m m, m m V-)
O < L L L) rr ) \j
cJ Ln CI Un Un LI<
z a- o o oo z t