Discussion
This section will discuss the conditions under which group and condition differences were found, and the differences between the three baselines, as well as the difficulty with the post-training baseline. It discusses which correlations between enhancement scores and experiential scores lend support to Fehmi's attentional hypotheses and which do not, and the "right" and "left-mover" differences in these correlations. It elaborates on the possible patterns of correlations and covers such other factors as subject motivation and "within" versus "between-sense" measures of experience. lastly, it suggests an alternative explanation for the correlation of EEG with experiential scores and summarizes the conclusions.
main hypotheses
The primary question is:
1. Did subjects, overall, report a different experience in the enhancement condition (higher synchronous alpha) from in the stabilization condition (lower synchronous alpha)?
The other two major questions are:
2. Did subjects report more of a sense of "merging" with their experience?
3. did they report a "widening" of attentional focus along several dimensions?
The answer to the first question is, as a group, no (see table 7). This hypothesis was supported only by the "left-movers" group, who did report different experiences overall but not on any dimensional subscale (see table 10). No test between means revealed any significant difference between conditions on any of the individual experiential subscales (e.g. "merging").
One possible explanation for this lack of significant difference between conditions on individual subscales and for the whole group on the experiential measure as a whole, was that subjects, overall, never raised their average EEG training scores above the grouped baseline for "A" condition, and hence, never achieved a clearly distinguishable "state" that may be associated with higher alpha enhancement levels. (subjects did however, raise their scores above all baselines, except the post-training baseline).
In the studies done by Hardt aND kamiya (1970, 1976, 1978) that successfully got significant changes in subjective ratings in relationship to EEG alpha amplitude, the subjects achieved clear, above baseline EEG amplitude scores. The present study, however, uses a different EEG measure than simple alpha amplitude. No well-controlled studies have been done using this measure, showing that above baseline amplitude levels are possible or are more easy or difficult to achieve. Since the current measures require both amplitude and synchrony increases over a larger area of the cortex, above baseline levels may be harder to achieve. Also, more time than five hours training may have been needed. Hardt (1975a) states that changes in occipital alpha are more likely to occur after six to ten hours. As previously stated, Fehmi, in his own learning to produce synchronous alpha, required over ten hours to begin to "control" his synchronous alpha.
Another possible reason for the entire group not having significantly different experiential scores between conditions is that the questionnaires did not adequately measure the relevant subjective dimensions. Fehmi's study (1974) used 45 semantic differential scales which measured self-perceived trait changes instead of the actual changes in experience during training that the present study uses.
Another possibility is that the biofeedback tone was not conducive to alpha enhancement. Tyson (1982) demonstrated that certain biofeedback tones, found unpleasant by the subjects, are very inhibiting on alpha production. The present study, however, numerous ratings did not find one negative rating of the experience of the biofeedback tone. It thus does not seem likely that Tyson's findings apply in this case.
The usual clinical training done by Fehmi and others, using this type of phase-sensitive EEG, involves the simultaneous use of a biofeedback tone and a strobe light set at a certain amplitude threshold and which flashes at the same frequency as the subjects synchronous alpha. The strobe light was not used in the present study due to its well-known ability to drive brain waves. This possibility would have totally confounded the EEG results in the stabilization condition.
Part of the explanation seems to be, that since only "left-movers" had significant differences in their enhancement scores between conditions, only "left-movers" had a significant difference in their experiential scores. This implies that only the changes in enhancement scores caused corresponding changes in subjective experience. The fact that "left-movers" had a significant difference is corroborated by Fehmi's study that found that "left-movers" were better able to increase their alpha amplitude while "right-movers" were better able to suppress it (Fehmi 1974).
experiential subscales
No between-condition differences are found in any experiential subscales (e.g., concentration) for either "right" or "left-movers" until one looks at the correlations between enhancement scores and experiential scores. When baselines are controlled for, all significant correlations for the whole group are for "A" enhancement scores. In other words, only when the subjects were "attempting" to increase their synchronous alpha amplitude, did changes in amplitudes correlate with the rating of their experience. In the "B" condition, in which subjects were only supposed to keep their synchronous alpha near baseline, one would expect less change in subjective experience if such changes in experience do correlate with increases in amplitudes as hypothesized.
baselines
One difficulty in subjects getting above-baseline synchronous alpha was caused by the inclusion of the third daily post-training baseline. This baseline tended to be higher than the other two and significantly raised the average of all three. This finding is also verified by Tyson's studies (tyson, personal communication, April, 1992). One possible explanation of a the higher third baseline is that the subject "let go" after making efforts to control the feedback and thus "rebounded" into a state of letting go of control. the experience "letting go" is traditionally associated with higher amplitude alpha. this possible "rebound" effect in line with the effect described by
fehmi (1974) where pilot subjects made effortful concentration on the biofeedback tone until they "let go" and reported "rebounding" into an effortless concentration. Unfortunately, most studies have not included a post-training baseline to study this possible effect. Future studies should include post-training baselines to measure this possible effect and take it into account.
Another uncommon measure was the use of baseline maximum scores (the highest amplitude one-minute epoch of each ten or five minute baseline). Neither group of training scores ("AC" or "BC") exceeded the "max" scores, as a whole, or for groups. Thus Plotkin's (1979) more stringent criterion for exceeding baseline amplitude was not met. Although not directly related to the hypotheses, an interesting finding in the use of maximum scores, was that they correlated more highly with reported subjective experience than average baseline or enhancement scores. Amplitude maximum ranges may thus be a more sensitive measure or correlate of subjective experience. This finding is somewhat in line with Tyson's (1979) finding that the variability of alpha is more highly correlated with experiential scores. Maximum scores are more likely to reflect alpha variability.
correlations with baselines
It appears that, for many dimensions, the subject's baseline amplitude on a certain day is related to how he will report his experiences of doing the training on that day. Overall, the higher the subject's baseline amplitude was, the less he was likely to report feeling "separate" when describing his experience of lower alpha (stabilization around baseline) during the training. Also the less likely he was to report higher scores on seven subscales of attentional widening (bwthnssi, bwthnsfs, btwnssi, btwnsfs, bmultisp, bfragint, bovrawid). When describing his experience of higher alpha (enhancement) the less likely he was to report higher score on only three subscale of attentional widening (awthnssi, awthnsfs, abtenssi) (see table 21). These findings are hard to interpret since the subject is not reporting on his experience of his synchronous alpha during baseline but rather during training. At present no explanation of these findings is offered and they appear to neither ad to, or subtract from, the support of the main hypotheses. Correlations with baselines are only partly focused on in the present study since they are not directly related to the original hypotheses.
The numerous correlations of experiential measures (including standard covariates) with the baselines, suggests that subject experiences during the training is actually partly a function of the subject's brain waves when he was just sitting, doing nothing, prior to and after the training (as during the baselines). The future implications of this relationship might be to use measures of baseline amplitude or synchrony as predictors of performance on other kinds of psychological self-assessment measures, or even of performance on certain other psychological tasks.
using baselines as covariates
Baselines were originally not going to be used as covariates because it was thought there would not be this strong relationship with experiential scores. The finding of a relationship between baseline and subjective experience may be one of the most significant finding in the study. To my knowledge, except for tyson (1979) and Hardt (personal
communication, April 1992), the great majority of other alpha biofeedback studies have not found this relationship or taken it into account. It would be extremely important to take these types of correlations with baselines into account especially when one is correlating EEG training scores with subjective reports.
Because of the correlations with baselines, it became apparent that variance associated with baselines scores would have to be partialled out if one were to find the effects of the training conditions on reported experience that were completely independent of baselines. As a result, many significant correlations disappeared when baselines were controlled for. Also, in some of the correlations, baselines and the standard covariates were used together as covariates to use to control for the most possible influences.
Possible patterns of correlations with Enhancement scores
The correlations mentioned above, along with other correlations with the entire group, "left-movers" and "right-movers," show a possible pattern for the two groups and the entire group. First of all, when baselines were controlled for, for the group as a whole, all significant correlations were with "A" enhancement scores (see table 25). In other words, only when subjects were attempting to increase the feedback tone associated with increases alpha synchrony and amplitude did they have correlated changes in experience. It appears that the increase in amplitude in the enhancement condition is related to changes in subjective experience. It is presently not clear if this is a causal relationship. The exclusive occurrence of significant correlations with "A" condition (enhancement) makes it much more likely that increasing ones alpha amplitude causes the changes in the subjective dimensions specified in the hypotheses.
support for the "Merging" hypothesis Were the above changes, however, in the predicted direction? The answer is yes for the first ("merging") hypothesis. The first hypothesis predicts a decreased subjective sense of being separate from ones experience in "A" condition. This hypothesis is supported by the fact that when baselines were controlled for, "A" enhancement scores significantly correlated negatively with "separation" for "A" experiential scores and only "A" experiential scores (p < .01) (see table 25).
Support for the "attentional widening" hypothesis hypothesis number two predicts that subjects will experience more attentional "widening". This would mean an increase in the "incorporation of opposite states" (e.g. experiencing trust and distrust) and an increase in "multi-sensory perception" in "A" condition (discussed later). this hypothesis is not supported in terms of "incorporation of opposite states" ("incorop") because "A" and "B" experiential scores for "incorop" correlated negatively (not always significantly) with enhancement scores in all conditions. The "A" (higher alpha) experiential scores for "incorop" significantly correlated negatively with "A" which is the opposite of what was predicted (see table 25). "B" (lower alpha) experiential scores had no significant negative correlation with "B enhancement scores.
"crossover" correlations with enhancement scores
"B" experiential scores for "incorop" also correlated negatively with "A" condition enhancement scores. This means that the subjects' assessment of condition "B" (their experience of "close-to-baseline" or lower alpha) did not correlate with condition "B" but rather correlated significantly with the enhancement scores of the condition they were not assessing ("a"). In other words, to some degree, a subject's experience in one condition may be indirectly related to, or influenced by, his brain waves in the other condition. The present study labels this type of relationship as "crossover." Perhaps the experience in one condition carries over to influence the subject's experience, or the memory of his experience, in the other. At present, there is no clear explanation for this "cross-over" effect. It does occur frequently in the present study (see tables 25,26 and 28).
another significant "crossover" effect occurred when experiential score were correlated with enhancement scores without controlling for covariates: As EEG enhancement scores increased in condition "A", subjects rated their "concentration" (narrowness of attentional focus) as lower for condition "A" (higher alpha) and also rated their concentration as higher for their experience of condition "B" (lower alpha) (supporting the attentional widening hypotheses in this dimension) (see table 23). Since baselines were not controlled for this finding is inconclusive.
when baselines were controlled for, "B" experiential scores for "multisensory perception" correlated negatively with "A" enhancement scores. this means that the greater EEG amplitude was for condition "A" the more subjects reported a decrease in "multisensory perception" for their experience of "close-to-baseline" or lower alpha in condition "B". This finding could remotely support the "attentional widening" hypothesis for "A" condition in that, if greater alpha amplitude is associated with an increased sense of "multisensory perception" then, possibly by some kind of perceptual contrast, one might experience less of "multisensory perception" in the "B" condition, the more one experienced it in the "A" condition.
It also seems reasonable that one may be better able to distinguish what his experience of "close-to-baseline" alpha is (condition B) after he has reported his experience of higher amplitude alpha (condition "A") because one can contrast one experience with the other. Of course this is only one possibility.
patterns of "left" and "right-movers"
When "left-movers" were looked at, the only significant correlations between experiential scores and enhancement scores were the "crossover" type discussed above, where EEG amplitudes in "A" condition were only associated with the reported experience of the "B" condition (see table 26). These correlations include "within-sense field size" and "multi-sensory perception" which was just discussed. This pattern does not seem to be explained or clarified by any of Fehmi's findings or theories.
When the correlations included both the baselines and the "standard covariates" as covariates, "within-sense simultaneous" and "between-sense simultaneous" were added to the above list of negative "crossover" correlations with enhancement score. In addition, "B" experiential scores for "between-sense simultaneous" correlated negatively with "B" enhancement scores which lends support to the hypothesis for decreased overall widening in condition "B" (see table
19).
"right-movers"
For "right-movers" the only significant correlations were between "A" experiential scores and "A" enhancement scores. "Within-sense field size" correlated positively with enhancement scores (one of the few significant positive correlations) (see table 27). This generally supports the hypothesis for attentional widening, however, only for "right-movers." according to Fehmi, however, these "right-movers" should show less, not more, positive correlation on this scale than "left-movers." "incorporation of opposite states" correlated negatively with "A" enhancement scores which, again, is the opposite of what was predicted. Overall, for "right-movers" evidence for support of any of the hypotheses is inconclusive.
"Standard covariates"
Because most of these correlations are all of specific experiential dimensions with actual changes in EEG amplitude, it is hard to attribute differences in experience between conditions "A" and "B" to factors like, overall differences in how subjects experienced the actual sound characteristics of the tone between "A" and "B" (Tyson, 1982), or to overall differences in the perceived task difficulty, frustration, abundance of feedback and perceived success during the training. Many of these above-mentioned correlations still stood even when these factors were used as covariates.
It was interesting that the covariates themselves had some significant correlations with the baselines and with enhancement scores. Only the covariate "arousal" (awakeness) significantly correlated with "A" enhancement scores after baselines were controlled for (see table 38). This confirms findings that a certain amount of wakefulness is necessary to produce higher alpha (plotkin 1979).
Plotkin's criticisms of past studies related to the present study
The standard covariates were also used in the manovas which tested for between-group and between-condition experiences. Plotkin's (1979) criticism of previous studies, that experiential reports during alpha enhancement are due almost entirely to expectations, sensory deprivation, introspective sensitization and perceived success can, for the most part, be dismissed in the present study. Enhancement and stabilization conditions were equated in terms of expectation, sensory deprivation and introspective sensitization. Subjects sat in the same room for the same amount of time for both conditions and received almost identical instructions and preparations in a double-blind situation.
In addition, the subject gave ratings of how much he attributed his experience to sensory deprivation. There were no between-condition differences for the ratings of sensory deprivation, the experience of the abundance of feedback, or of the use of an oculomotor strategy.
task difficulty for conditions "A" and "B" were kept equal as much as possible, however, subjects rated "perceived success" higher for condition "A" (mean score for "A" = 5.26, Std. dev. 1.71, "B" = 4.80, Std. dev. 1.08). also "task difficulty" was rated higher for condition "B" (mean score for "A" = 4.39, Std. dev. .99, "B" = 5.12 Std. dev. 1.05). These differences between conditions are less than one point on a nine-point scale. of course, theses scores on perceived success, task difficulty, sensory deprivation, oculomotor strategy were used as
covariates in the manovas and some of the correlations.
The variance due to these covariates seem to be adequately controlled for, and between-condition differences were still found for experiential ratings. It is thus hard to attribute these differences to factors like perceived success, or task difficulty as plotkin has pointed to in past studies.
"Constant (pre-baseline) factors"
"expectation-motivation" correlated positively with one enhancement score for condition "B" ("maxave") (see table 31). This shows that expectation and motivation may not only be related to how subjects rate their experience, but is also related to the amplitude of their eeg during training. This result, if valid, could possibly confound interpretations of experiential questionnaire scores in relation to expectation and motivation, since some differences in experiential scores could be either due directly to expectation and motivation, or due indirectly to the change in EEG caused by expectation and motivation. An interpretation of this result is hard to make since it was for condition "B", where subjects were supposed to not increase their synchronous alpha.
within vs. between-sense measures of experience In terms of significant results, there was no detectable pattern to distinguish within-sense measures from between-sense measures. One measure did not get more significant correlations than the other.
Questionnaires not focused on Questionnaires "pci", "E-2", "FR", and "post-study" were not completely focused on in the present study, because they were not directly relevant to the hypotheses, and partly due to constraints on time and resources.
possible alternative explanations to learning to increase EEG synchrony and amplitude
"left-movers" learned to produce higher enhancement scores in condition "A" than in condition "B" and all subjects learned to produce average synchronous alpha levels higher than the first two baselines. It appears that some learning to increase synchronous alpha did occur. Learning to increase synchronous alpha relative to baseline, however, may not be necessary to learn an experiential state which may be associated with elevated amplitudes.
In this study, the biofeedback tone "chirped" with the frequency and volume corresponding to the frequency and volume of the subject's "summed" synchronous alpha during the enhancement condition. In the stabilization condition, volume corresponded to a closeness to a certain amplitude point set near the daily baseline. In the enhancement condition, subjects could be aware of the exact moments they were producing higher amplitude synchronous alpha and they could be aware of their alpha's relative amplitude, whether or not they were learning to increase it.
Theoretically, they could learn to discriminate a mental state that may be associated with the exact moments of alpha bursts. They could report what those moments of synchronous alpha bursts were like on the questionnaire. Subjects were just asked to answer the
questionnaires in regards to what their experience was like when the feedback was at its greatest abundance.
The possibility also exists that, what the subject learned was a combination of learning this kind of discrimination, not dependent on learning control, and the learning of some additional control or mental strategy to increase the amplitude. Future studies might be designed, using this EEG measure, to determine how much a subject learns purely to discriminate and how much a subject learned to control any mental state associated with the amplitude increase.
subject motivation
A factor that might have inhibited the subjects learning to produce clear, above-baseline synchronous alpha, was subject motivation which, in regards to filling out the questionnaires, tended to be low. Most subjects found the questionnaires too long and tedious, and this might have affected their motivation toward the actual EEG training. expectation and motivation prior to training did significantly correlate with one of the "B" enhancement scores (BC/ maxav) (see table 31).
About half the potential subjects dropped out after two or three sessions due to reported time constraints or dislike of the questionnaires. Such long questionnaires, of course, are not recommended for future research. Only a few questionnaires, however, were not completely filled out. These were usually from the questionnaire "T", given at the end, which measured the standard covariate dimensions. On the manovas, usually about 40 cases out of 260 were rejected by the SPSS program due to missing data. Other tests were done to include the data from the rejected cases, and these tests showed no overall change in the significant results from including or not including the rejected cases.
No measure of the "person factor", [as put forth by Taub and school (1978)], was included in the study to assess its possible effect on motivation. An optimum "person factor", in term of attempting warm, friendly contact with the subjects, was maintained in the study.
conclusions
Why most of the correlations are negative (roughly two-thirds) is uncertain. This could be due to the nature of the questions, coincidence, or due to an unknown artifact. Most of the dimensions for the questionnaire E-1 measured "widening" of attentional focus, in one way or another. Higher eeg scores tended to be associated with less widening. Most of the attentional "widening" measures thus had negative correlations. The other two dimensions pertaining to the main hypotheses, were "merging" and "separating" which are polar opposites. "merging" tended to have positive correlations (although not significant) while "separating" tended to have negative correlations. because a positive and negative correlation was found for polar opposites, it thus appears that the predominance of negative correlations for "attentional widening" may be due more to the nature of the questions.
In summary, correlational support of Fehmi's hypotheses for the entire group, "left-movers" and for "right-movers", in most cases, tends to be sporadic and not be clear cut. Correlations of attentional widening experiential scores for "A" with enhancement scores for "A" tend to be negative which are in the "wrong" direction (see tables 23,
25,27, and 28) the fact that only "left-movers" showed an overall significant difference in reported experience between conditions and only "left-movers" showed a significant difference in enhancement scores between conditions, lends strong support to the most general hypothesis that there are a distinguishable attentional states associated with the learned production or inhibition of this summed amplitude signal.
The fact that the "separate" dimension significantly correlated negatively with only "A" enhancement scores offers good support for the "merging" hypothesis.
A general tentative conclusion could be made that, overall, the two basic dimensions the hypotheses looking at, "merging" and "attentional widening," may indeed correlate with EEG amplitudes, but the correlations with "widening" tend, in general, to be opposite of the direction predicted. in terms of predicted directions, suspicion was also raised about Fehmi's hypotheses when, at one time, he himself was unable to determine that he was receiving feedback contingent upon decreases in synchronous alpha when told he was receiving feedback for increases in synchronous alpha over an approximately five to ten minute period (personal communication, 1987).
"Right-movers" and "left-movers" showed different possible patterns of correlations. What these patterns may mean is yet unclear. Since "right-mover's" enhancement scores between conditions were not significant and those of "left-movers" were, the differences between "right" and "left-movers" in correlational scores may be partly attributed to the difference between learning and not learning to increase enhancement scores. The correlation with the "PCI" shows that some of these patterns are not just peculiar to the E-1 questionnaire.
What is clear is that there are processes involving eye-movement preference, baselines, maximum scores, etc. for which new hypotheses have to be proposed. In addition, if subjects can be trained using this type of phase-sensitive biofeedback equipment to clearly increase their synchronous alpha above post-training baseline averages or above baseline one or two-minute maximums, then perhaps different, or more clear-cut results might occur. It should be noted that most studies have not used post-training baselines as a reference, so if the present study uses the same type of baseline references as many past studies, then above-baseline enhancement scores were achieved by the entire group.
The present research has attempted to improve upon past research in several ways. It has looked at many different studies and has tried to look at their pitfalls and suggestions. Many of the measures and controls have been used in other studies, however, the present study has tried to assemble what is believed to be a unique and more comprehensive combination of measures and controls. In summary, these measures and controls are: 1. the use of much longer training times (five hours per condition), 2. the use of subject ratings on important covariates such as perceived success, task frustration, experience of the feedback, expectation and sensory-deprivation, 3. the first use of the PCI with EEG measures, (PCI experiential dimensions had many significant correlations with baseline scores) 4. the use of EEG baselines as covariates in some cases, 5. the use of post-training baseline, which proved to be higher than the two pre-training baselines, 6. the use of baseline maximum scores, which tended to have higher correlations with
experiential scores , 7. the extensive use of correlation to find new connections between EEGs and reported experience, 8. the use of a biofeedback signal that is sensitive to "in-phasedness" and that involves all major lobes of the brain instead of just occipital, (virtually no published studies have been done attempting to connect experiential states with biofeedback signals derived from large areas of the brain), 9. the first use of a control condition which uses "stabilization" of EEG, which has been found to be a viable alternate to other control conditions, like giving false feedback, which may have more pitfalls (strayer, Scott,& Bakan, 1973), (some of Fehmi's earlier research had no control group or condition at all), 10. lastly, the use of questions to comprehensively rate Fehmi's basic attentional variables ("merging" vs. "separating", "widening" vs. "narrowly focusing"). despite their drawbacks, the experiential questionnaires used showed that subjects can successfully be asked to make subtle discriminations in their subjective experience. These questionnaires uniquely allowed for the reporting of opposite states by avoiding the forcing of subjects to choose between two opposite experiential poles (e.g., choosing between relaxation and tension). "Incorporation of opposite states" was found to be a significant correlate of enhancement scores for "right-movers".
In general, present research extends past research in the sense that it has more strongly confirmed, in a more highly controlled way, that there are subtle attentional connections between EEG alpha and personal experience that go much farther than Plotkin's conclusion that higher alpha just means that one is awake and "not looking" (Plotkin 1979). These connections cannot be attributed to attending to "non-sensory" mental activities to avoid "looking" oculomotor activities as plotkin found in his studies, since correlations in the present study showed no such pattern.
Future research, looking for subjective connections with EEG processes, should look for experiential correlations with baselines and control for these baselines in some way when looking for correlations with experiential ratings.
other important factors which should be considered in future research are eye-movement preference, the use of post-training baselines, the inclusion of maximum scores in correlations, and the use of questionnaires which measure important covariates such as perceived success and task difficulty. The use of post-training baselines should be looked at in light of the possibility that subjects may rebound into more "letting go" and higher alpha after attempting to effortfully control alpha (tyson, personal communication, April, 1992) (fehmi, 1974).
The study of EEG and phenomenological scores, and of EEG biofeedback and phenomenological scores, although very limited for the past ten years, after becoming somewhat popular in the 60's and early seventies (Echenhofer & Coombs, 1987), is showing much more promise in terms EEG technology improvements. now possible, is the use of real-time "brain mapping" and other kinds of new biofeedback displays which can "feedback" many different combinations of EEG parameters (lexicor, 1990) (williamson, 1990). There are a great many different brain processes (e.g. certain topological distributions of frequencies, amplitudes and phase relationships) which could eventually be monitored
and used as feedback, some combination of which may be much more related to the higher order mental processes involved in the awareness of thoughts, sensations and feelings, than the relatively crude measure used in the present study. As computing speed and power increases along with the eventual development of artificial intelligence in the years ahead, such physiological measurements may become more possible (Tyson, personal communication, April, 1992). The use of adequate control conditions, a measure of the appropriate covariates, and some of the other factors illuminated by the present study may be very important in future research.
Clinically, studies by Peniston and Kulkosky (1989) using an alpha-theta type of EEG biofeedback (which also involved phase-synchrony) are currently showing promise in the treatment of chronic alcoholism (1989) and post-truamatic stress disorder (1991). Large gaps in the knowledge of EEG biofeedback still exist. As far as using general EEG alpha biofeedback as a treatment for stress, which is not focused on in the present study, Tyson (1987) concludes that considerable biofeedback research is needed and that "As Mulholland (1984) explicitly illustrates, the gaps in our knowledge justifying the use of EEG feedback as a rational therapy for stress and anxiety are much wider than the treatment of paresis with integrated EMG or Raynaud's disease with skin temperature, and these gaps will not be filled in the near future if the declining interest in EEG alpha biofeedback persists."