3. BIOLOGY AND POWERLINE EMF HEALTH HAZARDS
The biological studies consistently show that powerline EMFs can be detected by exposed subjects. For this reason alone, powerline EMFs should be presumed to affect human health.
There are two scientific methods for establishing scientific facts (Section 1). In principle, therefore, there are two ways in which scientific facts could be established that bear on the question whether powerline EMFs affect human health. The method of physics does not result in facts that materially support either side of the issue (Section 2). Here, I again consider the question of powerline EMF hazards, but in the context of the more general thought-style of biology.
Many disparate views regarding whether powerline EMFs affect human health have been expressed in editorials, informational pamphlets, government reports, journal articles, and books. The opinions differed even though the investigators who performed EMF bioeffects studies professed common goals for their experiments, and even those who offered global analyses all evaluated the same laboratory data.
Why do divergent opinions abound regarding the public-health significance of the EMF biological studies? My first goal is to show that differences in the hypotheses, norms, and theories of both the laboratory investigators and the expert reviewers caused the split in opinion. Different scientists did not reason the same way, and it is therefore not surprising that they reached different conclusions.
Because differences in biological reasoning lead to opposite conclusions regarding whether powerline EMFs affect human health, it is necessary to choose how the issue ought to be decided. My second goal is to explain why this decision rests only partially with scientists. It is the right of the public to decide some pertinent issues as, for example, the level of certainty to be used when evaluating the scientific evidence for the purpose of making policy decisions that affect public health.
Biological evidence about the effects of powerline EMFs can come only from studies in which animals or human subjects were exposed to electromagnetic fields and then observed to determine the consequences of the exposure. We expect that if it is true that powerline EMFs can affect human health, then some kind of a consistent pattern of changes will be observed in such studies. We recognize that the mechanisms may be obscure or even completely unknown, but we require, at a minimum, the existence of some reproducible or reliable phenomena that can serve as the basis of an inference that powerline EMFs can affect human health. Otherwise, we must conclude that no known evidence exists to support that inference.
The reported EMF bioeffects studies, however, appear to be highly problematical for at least two reasons. First, there are instances in which investigators failed to find an effect due to EMF exposure. For example, a group of investigators tested the hypothesis that exposure of lambs to powerline EMFs would alter melatonin patterns and thereby cause a delay in the onset of puberty. But lambs who lived under a 500kV powerline for 10 months did not exhibit detectable changes in serum melatonin patterns or onset of puberty. The investigators repeated the experiment, again with negative results, and argued that the studies were evidence against the theory that EMFs affect melatonin, which was a conclusion reached by other investigators who used different experimental designs. Whether or not it is justifiable, it is a fact that all EMF studies are viewed by some as dubious largely because of comparisons between negative and positive studies in which a particular parameter was measured using different experimental designs.
A second reason for uncertainty regarding the implications of the EMF bioeffects studies is that there appear to be inconsistencies involving similar experimental designs within virtually every line of EMF biological research. A pattern has emerged during the last 25 years in which a report of an EMF bioeffect in a particular animal model observed under particular conditions was followed by a second report by another group of investigators who performed a similar study but could not confirm the original results. This pattern has been repeated many times. Calcium adsorbed on brain tissue was reported released at different rates depending on the presence or absence of weak EMFs (1), but others were unable to reproduce this effect (2). EMFs affected skeletal growth in chicks (3), but the same model system did not yield positive effects in the hands of other investigators (4). Sometimes EMFs affected growth rate of animals (5), but not in other cases (6). EMFs altered transcription (7) or not (8) in seemingly identical experiments performed by different investigators. EMFs were or were not associated with cancer (9,10), affected or did not affect melatonin levels in the blood (11,12), and did or did not induce a stress reaction (13,14), modify behavior (15,16) or affect cell growth in vitro (17,18), again depending on who conducted and evaluated the experiment.
The inter-experimental-design species of inconsistency (the species of inconsistency resulting from different designs) is not important for the simple reason that it takes no skill whatever to design and perform a study that finds nothing. I will not deal with this issue here, but will treat it in a later section dealing with trade-association science.
The issue I want to address here involves the serious kind of inconsistency that apparently occurred when a group of investigators used an experimental design similar to that of an initial group but failed to find the same results. If the reality is that the exposed subjects did not detect the presence of the EMF, then the reports that failed to find a biological effect due to EMF exposure would reflect the objective state of nature. In that event, the positive reports would be artifacts, errors, or statistical fluctuations. It is crucial, therefore, to determine whether the results of the intra-experimental-design studies were actually inconsistent.
Possible Bases of Apparent Inconsistency
Early in the evolution of the dispute regarding whether powerline EMFs affect human health, some literature dealing with the issue was pregnant with the notion that essentially all positive reports were somehow due to poor experimental procedures on the part of the investigators. The criticism initially appeared as a series of accusations against Soviet scientists, and then spread to American and European investigators who reported EMF effects. Ultimately, however, as the EMF health-risk dispute developed it became broadly obvious that this explanation was baseless and inaccurate.
A second possible explanation for the apparent inconsistencies was that they resulted from statistical fluctuations. In this view, a few studies that looked positive were to be expected on the basis of statistical fluctuations alone. A difficulty with this argument was that each of the EMF studies was independent in the statistical sense, and each was protected at the 5% level against the statistical error of declaring an effect when none actually existed. Consequently, assuming statistical fluctuations were important, there was no reason to conclude that it was the statistical fluctuations associated with the positive studies that were misleading, rather than the statistical fluctuations associated with the negative studies. But even if the statistical-fluctuations argument was a good one, it applied only where a few kinds of EMF studies were performed. The argument failed to explain why putative statistical fluctuations occur in the context of every experimental design in which a positive effect was reported.
A third potential basis for intra-experimental-design inconsistency was biological variability. The proponents of this view pointed to circadian rhythms, genetic differences between individuals, microenvironmental factors, and the complexity of the neuroregulatory and immunoregulatory systems of the body, and argued that interactions among these myriad variables, not the consequences of EMFs, produced the claimed differences between exposed and control animals. But this explanation cannot be correct because it too is improbable. If it were true that the many interacting variables caused inferential errors in the biological studies, then the overwhelmingly likely direction of the error would have been towards failing to recognize true effects, rather than towards failing to correctly accept results as negative. Thus the argument is premised correctly (biological variability), but the conclusion is wrong.
Another explanation is that the appearance of inconsistency arose because of differences in purpose or plan among the investigators who performed the EMF studies, as reflected in their hypotheses, norms, and theories. To understand how, in principle, such differences could account for the appearance of inconsistency between studies that were intended by the investigators to be similar to each other, consider (hypothetical, for now) studies dealing with the effects of powerline EMFs on the growth rate of animals. Let W stand for the average value of the weight of a group of animals in a study and V stand for the variance in the weight. The subscripts E and C will be used to designate the experimental and control groups, respectively.
The purpose or plan of an investigator is reflected in his hypothesis. Possible study hypotheses include:
Hypothesis No. 1: WE is greater than WC
Hypothesis No. 2: WE is less than WC
Hypothesis No. 3: WE is not equal to WC
Hypothesis No. 4: VE is greater than VC
Hypothesis No. 5: VE is less than VC
Hypothesis No. 6: VE is not equal to VC
Hypothesis No. 7: WE and VE are not equal to WC and VC
Suppose results supporting Hypothesis No. 1 and Hypothesis No. 2 were observed in two different studies. It could be argued (and has been argued) that the studies were inconsistent. In a sense the argument is correct because identical results were not observed in different experiments. But in another sense the results were consistent because both studies agreed that EMFs affected body weight - they differed only with regard to the direction of the change that was observed. Perhaps the thing that needs to be explained is why the two effects occurred, not why they occurred in opposite directions. Thus, the results are consistent or inconsistent depending upon one's attitude regarding the meaning of consistent.
Now consider an inconsistency between positive and negative reports, which is the classic case. This occurs when a study that tested Hypothesis No. 3 found results that supported it (that is, found that the average weight in the exposed animals was either greater or less than the corresponding weight in the control animals), but another similar experiment with the same hypothesis did not (that is, failed to reject the null hypothesis). In this case, the reports are inconsistent about their implications regarding the effect of EMF exposure on the average weight of the animals. The implication of the positive report would be that the EMF was somehow detected by the bodies of the exposed EMF animals, resulting in a change in the average body weight. The implication of the negative report would be that detection of EMFs did not occur because, if it had occurred, the results would not have been negative.
There is a possible state of nature regarding this case in which the implications of the positive and negative studies would actually be consistent with one another. Suppose in the study that was apparently negative on the basis of Hypothesis No. 3, the variance was viewed as the test statistic (Hypothesis No. 6) with the result that the study was positive (that is, the null hypothesis was rejected). The state of nature would be that the positive study was positive because W was altered by the EMF, and the study that was judged negative because W was not altered would actually be positive because V was altered. Thus, the studies would be consistent because both would imply that the EMF was detected by the body.
I now want to show that in the actual EMF growth-rate studies, the apparent inconsistencies disappear when the hypothesis, purpose, and plan of the investigators are considered.
Powerline EMFs and Growth Rate
In the 1970s, Richard Phillips and his colleagues at Battelle performed two apparently identical but independent experiments dealing with the effects of powerline EMFs on the growth rate of mice. In each experiment, three generations of mice were exposed essentially continuously to EMFs under conditions designed to avoid artifacts that they perceived to be responsible for earlier positive results in experiments performed by my colleagues and me.
The results of their first experiment showed that the average weight of both the male and the female mice were less than their corresponding controls (Table 1). In the second experiment the average weight of the male and female exposed mice were significantly greater than the corresponding controls (Table 2). The investigators averaged the results of the first experiment with those of the second experiment and concluded that the data provided no evidence that powerline EMFs can affect growth.
How were they able to justify averaging the results of two independent, statistically significant experiments to conclude that no effects were seen? It was done by assuming a linear model for the interaction between EMFs and tissue. The investigators assumed that differences observed in the weights of individual mice in the control group were due to random fluctuations, and that any effect due to an EMF would be linear. In this model, an effect due to the field must be consistent from animal to animal and from experiment to experiment, regardless of all factors or conditions other than those explicitly controlled. If, for example, the EMF produced an increase in the weight in one animal and a decrease in a second animal, that result would violate either the assumption that uncontrolled factors were unimportant, or the assumption that the response was deterministic. For this reason, when Phillips found that the EMF mice in the second experiment were not smaller than the controls, as was the case in their first experiment, he concluded that the absence of a consistent change in the average meant that there was no effect due to the EMFs.
The chain of reasoning in the Phillips study began with the assumption that a linear model governed any possible response of the mice to the EMF, and went as follows: because no consistent effects on the average weight of the exposed mice were found, there was no linear response, and therefore no response at all; consequently, the experiments furnished no evidence suggesting that the EMFs were detected by the body; because there was no evidence of detection, the study provided no evidence of possible health risks. The important point regarding this reasoning is that its validity is entirely dependent on the validity of a linear model. In this model, consistency of change in the average value of the weight is an absolute requirement.
When Phillips visited my laboratory in September 1976, I objected to his plan to assume a linear interaction model. Although Phillips' experimental procedures were similar to experiments performed by my colleagues and me, we did not assume a linear model in the evaluation of the data, and therefore did not require consistency in the average value of body weight as a pre-condition before concluding that the EMF caused an effect. Instead, we evaluated the data as planned comparisons to assess whether there was or was not a difference between the exposed and control groups at the ordinary level of scientific certainty (5%). Because we did not assume that the effects of EMFs would necessarily be linear in nature, the character required to be manifested by the data was not consistency in change in the average value, but rather consistency in the finding of a difference between the exposed and control groups in particular experiments. Our rationale was that this kind of consistency would justify a conclusion that the EMFs had been detected by the animal. It is plainly true that consistency in the mean is sufficient but not necessary to support this conclusion.
I would interpret Phillips' studies not the way he did, but rather the same way I interpreted my own studies. His data showed that powerline EMFs consistently affected the body weight of exposed animals, even though the effect could not be predicted in individual experiments.
The difference between Phillips and me regarding our interpretations of our powerline EMF studies on body weight in animals was related to our attitude regarding the public-health implications of our work. Phillips sought the strongest possible evidence regarding the biological effects of powerline EMFs - a consistent effect on the average value - and planned to deny the existence of any kind of lesser evidence. Had he found the type of evidence he sought, powerline EMFs would have been conclusively established as health risks and it would be unthinkable that the power industry would routinely carry out involuntary exposure to powerline EMFs. The position of EPRI and the power companies who sponsored Phillips' work was that until this kind of conclusive evidence had been obtained, the scientifically proper public-health strategy was to do nothing.
I never accepted the industry position, hence I thought Phillips' efforts were entirely misplaced. From my viewpoint, the conclusive evidence that Phillips sought might be impossible to obtain. There might be no such thing as a consistent effect on the mean of body weight or any other dependent variable in a powerline EMF study. That state of the evidence would not prove that EMFs don't cause human diseases. It would prove only that a conclusive demonstration of powerline-EMF health risks was not possible. Consequently, for public-health purposes, I thought the linear model was overkill. Consistency in the mean would have provided conclusive evidence; but consistency in change would be enough to warrant an inference of EMF detection, and that alone might justify the implication of health risk.
Change, as reflected in experimental data, is typically measured by the variance. Consequently, I analyzed the published EMF reports, other than the ones by Phillips or me, to assess whether they provided evidence that EMF exposure consistently resulted in change. I searched the literature for all studies that might plausibly be viewed as similar to the studies we conducted. I looked for studies that involved exposure of animals under laboratory conditions to power-frequency EMFs for long periods of time for the purpose of assessing the effect on body weight. I included every such study I could find that had analyzable data.
Some of the studies reported an effect of EMF exposure on the average weight, and some did not report such an effect. Juxtaposition of the latter reports with the positive reports was what gave credence to the idea that the EMF growth-rate studies were inconsistent, and hence not a proper basis for setting public-health policy. But when I analyzed these studies, I found that they manifested a consistent effect on change in weight (Table 3). The studies involving effects of EMFs on body weight were therefore consistent if the effect searched for was change rather than increase or decrease. Only if the added condition that the change always occur in the sample mean were added, could it be said that the studies were inconsistent. I prospectively tested and verified the idea that powerline EMFs are detected by the body as manifested in a change in growth, even though the EMFs do not result in a consistent change in the average body weight.
With regard to the studies involving the effects of powerline EMFs on body weight, therefore, if the hypotheses, purpose, and plan of the investigators is taken into account in evaluating the data from a reasonably similar series of animal studies, the implications of the studies are generally consistent in the sense that they indicate the existence of a cause-effect relationship between powerline EMFs and changes in body weight.
The Nonlinear Model and Consistency of EMF Bioeffects
If declining the assumption of a linear model generally leads to an explanation of intra-experimental-design inconsistency, then it ought to be possible to show that this is the case in other lines of research besides those involving body weight. The Henhouse studies are another group of similar experiments that can be evaluated for this purpose.
In 1982, Delgado and colleague reported that EMFs caused skeletal abnormalities in chicken embryos. The report led to follow-up studies, some of which confirmed the effect and some which did not. One proposed hypothesis to explain the apparent inconsistencies was that they were due to differences in the exposure systems used in the studies. If everyone used exactly the same apparatus and procedure, then consistent results might be obtained. The exposure systems were therefore rigorously standardized and similar experiments were carried out in three laboratories in the United States and three in Europe. The result was that significantly more defective embryos were found among the EMF-exposed eggs, even though that result was not obtained in each laboratory (Table 4).
The sponsors of the international cooperative effort that led to the data in Table 4 went to extraordinary lengths to insure that all of the participating investigators followed exactly the same experimental design and procedure. It is unlikely that this kind of inter-laboratory synchronization of experiments will be attempted again soon because of the high costs. Ironically, a line of argument subsequently developed holding that effects of EMFs on skeletal development in chicks is not important for the purposes of evaluating potential health hazards of EMFs, even though that was largely how the studies were initially justified. But even if this view were accepted, the Henhouse effort would still be important because, far better than could have been imagined, it revealed the role of normally uncontrolled variables in altering the manifestation of EMF transduction. This was also the real message of Phillips' growth-rate studies. If neither the Battelle investigators nor the Henhouse investigators could eliminate the impact of these factors, despite great efforts and the expenditure of millions of dollars, it is safe to conclude that they cannot be eliminated. The most parsimonious explanation for both studies, therefore, is that the biological systems were highly sensitive to initial conditions that were not - and could not be - controlled despite all reasonable efforts to do so. As I showed in the previous section, this is a fundamental, defining property of nonlinear systems.
The apparent intra-experimental-design inconsistencies in the studies involving the effect of powerline EMFs on cellular transcription can also be resolved on the same basis that afforded resolution of the apparent inconsistencies of the body-weight studies and the Henhouse studies. The case of apparent inconsistency in transcription studies began when Goodman and her colleagues reported that powerline magnetic fields affected cellular transcription. They did many different experiments and the reported effect of the EMFs was different under different circumstances. Goodman's studies elicited much interest because they suggested a link between the powerline EMF issue and orthodox molecular biology. However, Saffer and Thurston conducted similar studies and found results that they said refuted Goodman.
They focused on a particular set of conditions (57 mG, 20 minutes' exposure), and reasoned that either exactly the same data that Goodman observed under those conditions must be observed in their laboratory (irrespective of the myriad differences in other environmental factors between the two laboratories), or Goodman's inference that power-frequency magnetic fields can alter cellular transcription was wrong. When Saffer and Thurston measured the average amount of mRNA produced by cells, the results did not differ from the average of the controls. But the variance in their experimental data differed significantly from that of the controls, showing that the powerline EMFs were detected by the cells in their study, resulting in alterations in message for protein. This was exactly the conclusion reached by Goodman.
The apparent intra-experimental inconsistencies in calcium studies can also be resolved. In a series of studies Adey and colleagues, and others, reported that EMFs had a significant effect on Ca2+ in a system involving in vitro exposure of parts of animal brains to EMFs. These studies were the impetus for Albert and his colleagues who conducted a similar series of experiments. They compared the average value of Ca2+ in exposed and control dishes containing brain tissue, and found no consistent change in average value in a series of 7 experiments (Table 5). They interpreted this data to indicate that the EMF exposure had no significant effect on Ca2+, a conclusion that was apparently inconsistent with the findings of Adey and others. However, the data can be analyzed using the L test to assess whether EMF exposure caused any change in Ca2+. The results indicated that EMF exposure produced a statistically significant effect. The study was therefore consistent with the results of Adey and others if the plan to interpret the results is modified to allow nonlinear effects to be recognized.
Apparent inconsistencies have also been manifested in human studies. In 1966, Howard Friedman and Dr. Becker studied the effect of EMFs on the reaction time of human subjects. The subjects were instructed to press a key as quickly as possible after the appearance of a red light, and the results indicated that the EMF significantly affected reaction time performance. In 1995, Podd and colleagues repeated the experiment, and concluded that the EMF had no effect on reaction time. But even though the two studies were similar regarding exposure conditions and laboratory data acquisition, they differed markedly regarding their hypotheses and associated statistical designs. Friedman and Becker evaluated their data using an F test, to evaluate the effect of EMFs on variance. In contrast, Podd and colleagues used an ANOVA which entails an assumption of linearity. A true comparison, therefore, would require the use of the F test to evaluate Podd's data. When I did this, the result was that the implications of Podd's data were consistent with those of Friedman and Becker's data and showed that EMFs affected human reaction time (Table 6).
A final example of how the EMF bioeffects studies are consistent when the assumption of a linear model is avoided is provided by the work of Stern and colleagues. In two experiments, they said they found no evidence that EMFs disrupted the operant behavior of rats. This conclusion was opposite to that of Thomas and colleagues, whose experimental procedures were duplicated by Stern et al. But their data actually supported the conclusion of the earlier study (Table 7).
It is unnecessary to labor further regarding the point that intra-experimental-design inconsistency in EMF bioeffects studies is an artifact that results from differences between investigators regarding hypotheses, purposes, and plans to evaluate data. When apparently inconsistent studies were evaluated on a common basis, the inconsistencies disappeared. This was the result in each instance of apparent inconsistency that I analyzed. I expect that, ultimately, some exceptions will be identified, but it is difficult to imagine that they would amount to anything other than exceptions to the general rule. It can be concluded, therefore, that despite differences in models and statistical methods that were chosen and utilized by particular investigators in particular studies, the bottom line is that there is clear and convincing evidence that powerline EMFs were consistently detected by the various biological systems that were studied. It is simply not possible to gloss over the existence of this consistency.
Reproducibility of Nonlinear Phenomena
The conflict that Saffer and Thurston claimed was created by their results in relation to Goodman's results was apparent, not real, because it could be explained by taking into account the investigators' reasoning. The actual changes observed depended on the ionic composition of the solutions used, the temperature, the pH, the presence or absence of trace amounts of contaminants in the solution, the passage number of the cells, as well as many other factors, in addition to the field of 57 mG for 20 minutes. It is impossible to reproduce these conditions, and consequently it is impossible to reproduce specific changes in the average amount of expressed message. The same reasoning explains all the other cases of apparent inconsistency.
In general, the inability to precisely reproduce all conditions that can impact the biological system under study may or may not be a significant concern. If the phenomenon under study can be adequately explained on the basis of a linear model, then the consequences of the inability to precisely duplicate the laboratory conditions will be unimportant as long as the contribution to the variance in the dependent parameter due to the uncontrolled variables is less than the magnitude of the consistent effect caused by the independent variable. In this case, it is possible to replicate data between laboratories because the consequences of the differences between the laboratories are immaterial. But the situation is quite different if the linear model is not applicable, as in the case of powerline EMF bioeffects. In this case, small differences between conditions in different laboratories can have disproportionately large consequences. Because it is impossible to reproduce these conditions, it is impossible to reproduce the data.
One can decide that a nonlinear model is needed whenever intra-experimental-design inconsistencies inferred on the basis of a linear model can be resolved by eliminating the assumption of linearity. The consistency that is required to rationalize a judgment that a phenomenon exists is consistency in observation of the phenomenon, not consistency in the measurement of data (which is impossible for nonlinear phenomena).
Allowing the possibility that powerline-EMF bioeffects can be nonlinear does not entail that no EMF effects are linear. In other words, evidence of a nonlinear effect under one set of circumstances is not evidence against linearity under other circumstances. The best way to understand nonlinear is as the most inclusive term describing physical or biological systems. Nonlinear therefore includes linear, and linear is seen as a special case. For example, a pendulum is a nonlinear physical system that can be modeled as a linear system for situations involving small angular displacements.
As we have seen, the need for a nonlinear model can sometimes be manifested by employing statistical tests that involve comparisons of average values, but without the assumption of consistency in the average (which is equivalent to assuming a linear model). In other cases, applicability of the nonlinear model is manifested by employing statistical tests that involve comparisons of variance. In either case, if the underlying study hypothesis is accepted (null hypothesis rejected), then occurrence of detection of the EMF can be inferred. Because either statistic can be used to rationalize detection, the most sensitive experimental hypothesis ought to include them both, with appropriate protection against family-wise statistical error. One way this can be accomplished is by use of the L test.
Biological Generalizations Generally
The human-health implications of the fact that powerline EMFs can be detected by the body must be judged. That means all the evidence must be evaluated in some way according to some standard, because biological generalizations always require a framework of methods and standards. In this section I will show it is generally true that opinion, purpose, and values are important at this level of biological reasoning. In the next section, I will show that this is particularly true of the judgment regarding powerline-EMF health hazards.
Two hypothetical examples are sufficient to show the importance of subjective considerations in the formation of biological generalizations. First, consider the conclusion that decreased cyclin-E/CDK2 activity (Section 1, Table 1) causes loss of anchorage, which the authors suggested was generally true, based on their observations in KD cells. Assume that another group performed a similar study using XYZ cells, but did not find such a relationship. Is the abductive generalization suggested by the original authors now less reliable? If replicability were required, then the failure to confirm the initial results would cast doubt on their reliability. But failure to find something is not necessarily good evidence that the thing sought does not exist. Thus the hypothetical second report would not have proved that the phenomenon does not exist generally, just as it was the case that the first study did not prove that it does.
In practice, the attitude adopted toward such a mixed state of the evidence usually depends on the interests of the person or group deciding the significance of the mixed results. An author of a review article might hedge a decision ("the data is conflicting, and no firm conclusion is possible"). But there will be others who must take a position, perhaps because one conclusion or the other would materially influence the design of their experiments. Ordinarily, in resolving the question, many factors would be considered including known or suspected properties of the cells, degree of respect for the investigators, the reputation of their laboratories, whether the laboratories were in industry or academia, the track record of the investigators, insider information, style of presentation of the results, the relative prestige of the investigators' institutions, and perhaps even the nationality of the investigators. The point is that, in the face of mixed results, which is commonly the case, the cognitive value of the scientific evidence in a particular area depends on who is evaluating it, why he is doing so, and how he does it. There is no necessarily right or wrong means of performing these analyses.
As another example of the role of judgment in forming biological generalizations, consider the conclusion that vigilance caused an increase in brain blood flow (Section 1, Table 1). Assume that exactly the same change in blood flow occurred when subjects were exposed to powerline EMFs. To avoid the difficulty of mixed results that was just discussed, assume further that the study was replicated many times, and always with the same result. Would such evidence indicate that powerline EMFs would affect human health? Because a change in blood flow accompanies every cognitive act and every sensation, it could be argued that changes in brain blood flow caused by EMFs were normal physiological responses, and thus not hazardous. On the other hand, a change in blood flow also accompanies every pathological change and perhaps the rule should be that it would be better to err on the side of caution and tentatively regard the exposure as a hazard, at least in the cases where the exposure is involuntary. Thus two opposite conclusions are possible on these facts and again, the validity of the scientific inference depends on the reasoning principle chosen.
It can be seen that formation of scientific generalizations in the biological thought-style generally involves non-empirical elements, including opinion, purpose, and values. These elements are outcome-dispositive principles, and they cannot be chosen scientifically. Individual scientists differ in education, perspective, attitude, approach, experience, integrity, and ethical orientation. Disagreements can therefore be expected regarding how the biological thought-style ought to be implemented in a given case, for example, that of assessing whether it is a scientific fact that powerline EMFs affect human health.
The Generalization About Whether Powerline EMFs Affect Human Health
Suppose that a group of scientists were identified who shared a common set of scientific reasoning principles that, for example, included how certain kinds of measurements and observations should be made, how the data should be analyzed, assumptions deemed to be reasonable, and general laws. The principles provide a group with a frame of reference for deciding what should be accepted as scientific fact. When a group of scientists commonly accept a particular set of principles, I shall refer to them as a thought-group. Thought-groups may be large such as the groups consisting of radiation biologists, immunologists, microbiologists, or biochemists, or they may be small such as NIH study sections or blue-ribbon committees charged to decide whether powerline EMFs affect human health.
The investigators who performed EMF studies while employed at Battelle comprise a reasonably well-defined thought-group regarding EMF biology, because perusal of their estimated 500+ EMF publications and presentations indicates that they have a shared set of non-empirical principles. For example, they think that animal studies can be used to discern the existence of health risks to human beings. They think that mathematical modeling of EMF animal interactions can help determine the extent to which EMFs may be a health risk. They think that whether or not EMFs are presently recognizable as a health risk cannot presently be adequately assessed, and that therefore more research is needed. They regard the occurrence of linear dose-response relationships as an important relationship in ascertaining whether EMF effects in animal are real. These principles do not exhaust the shared reasoning principles among the Battelle investigators. They do indicate, however, that the Battelle investigators can be considered as a thought-group (Figure 1). No Battelle investigator has publicly opined that powerline EMFs affect human health. It is reasonable to infer that this result is a consequence of the particular principles that are shared by the group. Others who did not share these principles might not agree with the result.
Any ad hoc committee that interacts for the purpose of forming collective opinions necessarily defines a thought-group. For example, the experts chosen by the NIEHS to write a draft report for the NIEHS Working Group constituted such a group (Table 8). It would be improbable for the reasoning principles accepted by the NIEHS group to be identical to those of the Battelle investigators. Perhaps it is the case, for example, that the NIEHS group would require a different degree of certainty than would the Battelle investigators in assessing whether a given series of biological observations could properly be interpreted to indicate that powerline EMFs affect human health. Identifying differences in principles is possible, but that is not the point here. I want only to indicate that it is likely that some pertinent reasoning principles differ between the NIEHS and Battelle groups. If so, then the two groups will not agree on the factual status of some statements (see Figure 2). Whether or not powerline EMFs affect human health could be one such point of disagreement, depending on the consensus of principles adopted by the NIEHS committee. It is important to recognize that such a disagreement would not be based on data or measurements or observations, but rather on how the information was interpreted in the light of the axioms adopted.
In some instances, thought-groups are sharply defined because they were explicitly assembled on the basis of homogeneity of thought regarding a particular conclusion. Such was the case, for example, with the two groups of scientists who testified in a court proceedings in New York regarding whether powerline EMFs affect human health (Table 9). There was essentially no intra-group disagreement regarding the ultimate issue, but complete inter-group disagreement regarding it. The reason for the disagreement was the adoption by the two groups of materially different reasoning principles in evaluating the scientific data. Watson's group, for example, emphasized the absence of conclusive evidence, and the absence of known mechanisms, and the inability of Battelle investigators to replicate some biological effects reported by others. The landowners' witnesses, on the other hand, did not require that the evidence be conclusive, and largely rejected as irrelevant many of the concerns of Watson's experts. As a consequence of their choices, the two groups flatly disagreed regarding whether powerline EMFs affect human health.
A fourth example of a biological thought-group is provided by the Radiation Study Section of the National Institutes of Health. The hostility of this panel (and its predecessor) towards research proposals involving the study of nonionizing radiation is legendary in the EMF community. The attitude of the Radiation Study Section, however, is entirely consistent with the principles of radiological science espoused by the type of expert normally appointed to the panel.
The reasoning principles of radiation-panel experts can be inferred by considering the critiques they provided me regarding powerline EMF proposals that I submitted. Perusal of the critiques makes it clear, I think, that the radiation-panel experts have empirical reasoning principles that result in highly skeptical opinions regarding the existence and importance of EMF-induced bioeffects. It is unthinkable that the Radiation Study Section would accept the statement powerline EMFs affect human health as a scientific fact. The important point is that this result is a consequence of the opinions and values of the members of the Radiation Study Section thought-group, and does not follow in any scientific fashion from the biological evidence. The validity of their decisions is based on legal principles (they were duly appointed by somebody at NIH), not on scientific principles (there is no reason to believe that their opinions are objectively correct, or broadly acceptable to non-radiological scientists).
This analysis shows that a group judgment regarding whether powerline EMFs affect human health depends strictly on the opinions, purposes, and values that are commonly held by its members. Different groups hold different principles and, consequently, can be expected to make different judgments.
As best I can tell, there is no serious dispute (or no serious basis for a dispute) regarding the single most important scientific fact pertinent to deciding whether or not powerline EMFs affect human health. The important, resolved issue is that biological effects caused by electromagnetic fields of the type produced by powerlines actually exist. These effects are real. In the 1970s, this view was accepted by only a handful of scientists when the evidence for it was first marshaled by me. Today, however, it is the overwhelmingly dominant view among knowledgeable experts, and it is not possible to find a modern rational analysis that leads to a contrary conclusion.
The conclusion that biological effects due to electromagnetic fields actually exist is pivotal in the analysis of potential health hazards, and I hope the reader appreciates its significance. Were it the case that EMF bioeffects did not exist, then all of EMF biology would be a chimera having no meaning or significance within the framework of science. An assertion that powerline EMFs affect human health would, in that case, be entirely vacuous. On the other hand, because the available evidence clearly shows that EMF bioeffects do exist, it is as certain as anything in science can be that there exists a mechanism within the body that is capable of detecting and transducing electromagnetic fields into the language of biology - electrical changes in the nervous system, enzymatic activity, and protein expression.
The existence of EMF bioeffects and their necessary implication regarding mechanisms give rise to different kinds of issues. There exist scientific issues, which are in the domain of scientists. There also exist non-scientific issues which are properly in the domain of the layman (which, whether for reasons of arrogance or ignorance, have frequently been addressed by scientists).
It is a scientific issue whether EMF-induced changes actually occurred in cells or animals in particular studies. The further question regarding the choice of the model that best fits the data is also a question properly addressed by scientists. Elucidation of the biophysical principles that explain how the body detects powerline EMFs probably constitutes the most fundamental and difficult challenge to scientists. The rewards to humanity if we choose to fund an effort to meet this challenge are potentially great because we would gain information about ourselves, about how we work, as opposed to information about the nature of the planet or the structure of subatomic particles, as was obtained in other massive government science programs.
But the immediate question is not whether we have the political will to expend the money necessary to understand the electrical structure of our bodies. The question is what implications can properly be drawn from the presently available data regarding whether powerline EMFs affect human health. This question is not a scientific question because it cannot possibly be answered on the basis of laboratory data alone. It can be answered only on the basis of laboratory data and a set of rules that instruct the decision-maker regarding how and under what conditions the answer ought to be obtained. These rules are an indispensable aspect of generalizing from the biological data to make decisions about EMFs and public health. We need consider the situation involving only one rule, to understand the necessity of rules.
At least five qualitatively different standards for evaluating the evidence can be delineated. One possibility is that the evidence must be conclusive before the existence of a public-health risk is accepted. Conclusive would correspond to a standard such as beyond a reasonable doubt, or more than 99% certain. The typical scientific standard of 95% is another possibility. Perhaps the standard should be clear and convincing (75%) or a preponderance of the evidence (51%). Finally, it could be argued that a decision regarding whether powerline EMFs will affect human health should be made on the basis of an evaluation of the evidence in which the question is answered affirmatively if the evidence shows that such an effect is reasonably possible, say 25%.
My personal view is that 51% is certainly enough, and 25% may be enough. Others, I know, disagree profoundly with this opinion. Proponents of >99%, 95% and 75% can probably be identified. But whether you agree or disagree with my opinion that the standard should be at most 51%, it should be recognized that the choice is a sociological question not a scientific question. It is not the laws of science that dictate that the degree of certainty should be this percentage or that percentage, but rather it is the opinion of the larger society that properly sets the applicable norm.
I think it is clear that before deciding the substantive issue regarding whether powerline EMFs affect human health it is first necessary to decide what the rules of decision-making shall be. It is similarly clear that the choice of the applicable rules rests not with the narrow constituency of scientists, but with the larger society.
The fact that the biological evidence consistently shows that powerline EMFs are detected by the body raises the possibility that powerline EMFs affect human health. Whether this inference is acceptable is a sociological question not a scientific question because it can be resolved only by incorporating societal values, not by performing scientific studies. The essential societal value I would incorporate is the prohibition against involuntary human experimentation. The consequences of an erroneous decision are truly significant for the people who are involuntarily exposed to powerline EMFs, but relatively insignificant for the power companies. My personal sympathies lie with the involuntarily exposed resident along the powerline right-of-way, rather than with the power companies and their shareholders who would ultimately be required to pay the higher costs needed to design and build safer powerlines. I would therefore opt to protect the individual, rather than the power company or the aggregate of society. On this basis I would accept no higher than 51% certainty as sufficient. I think the scientific evidence meets this standard.
Marino Home Page | Research Interests
| 42971 |