Such horrors should be the stuff of nightmare or the merely metaphorical rancors of old prophecy. They aren't. They are the news of the day, by now even growing stale (for some) with reiteration.^[3]

And the respected historian of technology Lewis Mumford claimed in 1970 that

the professional bodies that should have been monitoring our technology . . . have been criminally negligent in anticipating or even reporting what has actually been taking place. . . . [Technological society is] a purely mechanical system whose processes can neither be redirected nor halted, that has no internal mechanism for warning of defects or correcting them.^[4]

Such opinions are extreme, but many other thoughtful observers of the human condition have expressed serious forebodings about environmental catastrophe. There is also another reason for concern, and that is that preventing an ecocatastrophe cannot always be accomplished by trial and error. Social scientists who study decision making generally agree that trial and error is a necessary component in human decision making about complex problems. However sophisticated the analysis and planning is that goes into a decision, uncertainties at the outset ordinarily are so great that errors are inevitable. It is by paying attention to these errors once they become apparent that individuals, organizations, and governments learn how to improve their choices.^[5] Unfortunately, the environmental problems that can lead to ecocatastrophe have several characteristics that make them unusually difficult to ameliorate with a trial-and-error approach.^[6]

The "trials" in some environmental policies often affect very large areas-even the entire globe-placing billions of people potentially at risk. Thus, errors in environmental policy can have potentially catastrophic consequences (for instance, a cancer epidemic) and waiting for errors to emerge before correcting policy therefore appears to be an unpromising regulatory strategy.

- 4 -

Moreover, when errors are made, they may result in consequences that are partially or wholly irreversible; such is the case with contaminated drinking water aquifers that take centuries to regenerate, the extinction of a species crucial to a significant food chain, and improper disposal of nuclear and toxic wastes that can affect humans and animals for very long periods.

In addition, because correcting an error requires that first an error and its consequences be recognized, the long lag time prior to feedback (that is, emergence of information about errors) can prevent for many years the realization that a problem even exists. It took some twenty-five years for persuasive evidence about the harmful effects of DDT to accumulate; evidence on lead and asbestos required even longer. It is not only that delayed feedback results in the accumulation of undesired consequences over the period required for the error to become apparent, but that, as a group of global modelers reports: "Owing to the momentum inherent in the world's physical and social processes, policy changes made soon are likely to have more impact with less effort than the same set of changes made later. By the time a problem is obvious to everyone, it is often too late to solve it."^[7]

Finally, there are so many potential sources of important problems that the sheer number may interfere with effective monitoring of emerging errors. For example, there are more than sixty thousand chemicals now commercially used in the United States. The result, according to a respected group of professional risk assessors, is that "if hazards are dealt with one at a time, many must be neglected. The instinctive response to this problem is to deal with problems in order of importance. Unfortunately, the information needed to establish priorities is not available; the collection of such data might itself swamp the system."^[8]

The predicament is that society is wrestling with risky technologies whose inherent nature makes them difficult to regulate by the traditional process of trial, error, and correction of error. Since error correction is the key to good decision making on complex issues, social scientists offer at least partial support on this particular point for the concerns of critics of environmental policy.

- 5 -

Recent Health and Safety Experience

Given the warnings of environmentalists and the theoretical concerns of social scientists, we would expect to see severe health and safety problems resulting from the risky technologies deployed increasingly during the twentieth century. In fact, health and safety trends to date do not match this expectation.^[9] Interpretations vary, of course, and there certainly is ample room for improvement in health and safety measures. But given the challenge posed by modern technologies, the record to date is surprisingly good: despite dire warnings, no catastrophes have occurred in the United States.

Civilian nuclear power provides a good example. Clearly it has not become an energy source "too cheap to meter," as one early advocate is said to have forecast. Moreover, the nuclear power issue is a source of tension and conflict in American society. Yet nuclear reactors have been operating for more than two decades without significant adverse effects on public health. This record is especially impressive since it was achieved in the context of a policy-making process centered in the Atomic Energy Commission that many now consider to have been badly flawed because of inadequate outside scrutiny.^[10]

By all accounts, Three Mile Island was the worst reactor mishap in the history of the American nuclear industry, and it was a financial disaster for the utility that owned the plant. But from a safety perspective, the accident was about the equivalent of a car accident. Some radiation was released, but the average dose to Pennsylvania residents within fifty miles of Three Mile Island (TMI) was less than 1 percent of the average background radiation to which they are exposed each year.^[11] The accident revealed a plethora of faults-poor management, maintenance errors, operator errors, and design errors (see chapter 3)-but even these did not lead to severe consequences for public health.

Widespread use of toxic chemicals also causes public alarm. The United States took the lead in introducing into the ecosystem hundreds of billions of pounds of inorganic and synthetic organic chemicals. At the outset we knew virtually nothing about the effects of these chemicals on human health, wildlife,

- 6 -

ocean life, microbial processes, and other aspects of the ecosystem. Ecocatastrophe might well have resulted from such unexamined and widespread use of toxic substances. To date, however, it has not. Because pesticides are designed to be toxic and are deliberately sprayed into the ecosystem, they provide an especially good example of a potential ecocatastrophe. By most standards, the average consumer in the United States has less to worry about from the effects of pesticides during the 1980s than in the 1960s or perhaps even in the 1940s.^[12] The same appears to be true in occupational and consumer exposures to toxic substances, although the data on this are currently inadequate.

But what about rising cancer rates? Have we already made millions of people ill by our toxic substances policies? There is no question that cancer is more frequent now than it was a century ago. But the change in cancer rates since synthetic chemicals came into common use is of a different nature than many realize. Lung cancer has increased rapidly, primarily from cigarette smoking. Digestive and cervical cancer have dropped significantly, and breast cancer death rates are stable. These types of cancer together account for about 75 percent of all cancer deaths. Rates of some other forms of cancer have declined slightly, and some have increased moderately; only a few types of cancer, such as mesothelioma (from asbestos) and melanoma (skin cancer), have increased dramatically. So there is no overall cancer epidemic, and some of the observed increase is due to increased longevity. In some occupations and in certain geographical areas, of course, exposures to asbestos or other dangerous substances have been much greater than the nationwide average, and higher local incidences of cancer have been the result.^[13]

To the extent that cancer is caused by environmental exposures, chemicals were the prime suspects in the popular media during the early and mid-1970s. Most cancer researchers, however, put industrial and consumer chemicals relatively far down on the list of carcinogens. Tobacco comes first, closely followed by the high-fat modern diet; alcohol and radiation (primarily sunlight and natural background radiation) rank next. Then in fifth place comes occupational and consumer

- 7 -

use of chemicals, along with air and water pollution; this amounts to 4 to 10 percent of cancer deaths according to most careful estimates.^[14] Much of this fraction is caused by such occupational hazards as asbestos, mining, welding, woodworking, and coke oven emissions. Some of these cancers involve chemicals, but they are not the synthetic organic chemicals that we ordinarily associate with the toxic chemicals problem. So the overall effects of synthetic chemicals to date have not been catastrophic.^[15]

Recombinant DNA (rDNA) research offers another example. During the mid-1970s, the prospect of splicing genetic material from one organism into another triggered fears of new and virulent epidemic diseases. Yet rDNA research has a perfect safety record to date; there have been thousands of experiments without any known health problems. However, the field is still in its early stages and is now moving from the laboratory into the environment where there are new risks. But experience to date is so reassuring that the burden of proof is now on those who believe this research involves unreasonable risks. Such critics are few, though some observers fear the release into the environment of newly created organisms (such as microbes that eat oil spills), and others are concerned about the increase in scale of recombinant procedures from small laboratory-scale experiments to industrial-size vats. Nevertheless, to date, the safety record is a good one.

It is possible to cite other examples, but by now the point should be clear: despite justifiable warnings and widespread public fears about the risks of complex new technologies, our society so far has escaped most of the predicted damage to human health.

A System for Averting Catastrophe?

How has this relatively good health and safety record been achieved? It is not the outcome many people expected. One possibility is that we simply have been lucky. Or perhaps the unhappy consequences from these technologies have not yet fully emerged (cancers from use of chemicals in the past

- 8 -

twenty years still could show up, for example). A third possibility is that idiosyncratic reactions to particular dangers have served to avoid the worst outcomes, but there has been no carryover of learning from one type of risk to another. If this is true, there is little assurance that the next danger will be averted. Finally, it is conceivable that the good record somehow is a systematic product of human actions-the result of a deliberate process by which risks are monitored, evaluated, and reduced.

The purpose of this volume is to explore this last possibility. In the chapters that follow, we examine the regulation of five types of technological risks: toxic chemicals, nuclear power, recombinant DNA research, threats to the ozone layer, and the greenhouse effect. In each chapter we will attempt to discern the strategies pursued in diagnosing and attempting to avert these threats.

At the outset of this research, we approached this subject with the commonly held assumption that the United States had botched the job of regulating risky technologies. Yet when we actually delved into how regulators have coped with the various risks, we discovered a surprisingly intelligent process. That is not to say the outcomes are fully satisfactory; but the strategies are far more sensible than we expected. The strategies we found usually were not fully developed. Nor were they always implemented effectively. And, in most of our cases, some useful strategies were ignored or underemphasized. But taken together, the strategies we found in use suggest the elements of a complete system for averting catastrophe. This system has five main elements.

First, an obvious early step is to protect against potential catastrophes. Part of this effort aims at preventing problems from occurring. But since the threats usually are too complex to fully envision and avoid, it is necessary to devise protections that limit the damage. We found considerable effort devoted to this goal in four of our five cases, although very different tactics are employed in each.

Because of the great uncertainty about the likelihood and magnitude of potential catastrophes, a second and related strategy is to proceed cautiously in protecting against potential

- 9 -

catastrophes. Decision makers can assume the worst rather than expect the likely. On this count, our cases display a mixed pattern. While there are important examples of such caution, we believe it would have been better to be more conservative about some of these risks.

A third strategy is to actively attempt to reduce uncertainty about the likelihood and magnitude of potential harm by testing whenever possible, rather than simply waiting to learn from experience via trial and error. Such testing is employed extensively in our cases and proves useful in preventing or limiting some potentially severe and irreversible errors.

Because the number of risks is too large to devote equal attention to all, a fourth step is to set priorities. Explicit procedures to achieve this end have been developed for some aspects of toxic chemicals, but less formal processes for priority setting are used for other types of risk. For at least one risky technology we studied, we found a clear opportunity for better priority setting than has been employed to date.

A fifth element is learning from experience, and we were surprised to find how much of this has occurred. In part because the ecosystem (so far) has been more forgiving than reasonably might be expected, learning from previous errors has been an important component of the system for averting catastrophe.

As we examine the regulatory histories of the five cases of technological risk, we consider each of these five strategies: protection against severe risks, erring on the side of caution, advance testing, priority setting, and learning from error. In each case our objective is to examine how potential catastrophes have been averted and to learn what strategies have been developed that can be applied to future problems.

Preview of the Analysis

Our inquiry begins in chapter 2 and concerns regulations governing the use of toxic chemicals. The questions considered are (1) How were toxic chemicals regulated prior to the emergence of public environmental consciousness?

- 10 -

(2) What efforts have been made since 1970 to improve on early regulatory strategies? (3) Do these new programs appear to be succeeding? This chapter shows the surprising effectiveness of trial-and-error as well as two other decision-making techniques that improve on this strategy.

Chapter 3 is about nuclear power. The potential for catastrophe here is strikingly different from that posed by toxic substances: the effects of a nuclear power plant accident would be highly concentrated. Therefore, the risk-reduction strategies that have evolved are at least superficially quite different from those for toxic substances. The chapter reviews safety procedures that evolved during the 1950s and 1960s; the Three Mile Island accident is then analyzed as a test case of those procedures.

The fourth chapter concerns an aspect of research in genetic engineering-recombinant DNA-in which DNA is taken from one organism and spliced into another. During the mid-1970s this procedure was as controversial as nuclear power; it was feared that some genetically altered bacteria would escape from laboratories and negatively impact human health or the environment. These and other concerns all but disappeared by the end of the decade, at least partly because of the development of rules for and safeguards against this possible hazard. This is our most definitive case, and in this chapter we come close to realizing the complete catastrophe-aversion system in operation.

Chapter 5 considers a threat to the global atmosphere-chemical assaults on the ozone layer that filters out a portion of ultraviolet light before it reaches the earth's surface. In 1970 several scientists suggested that flights by supersonic transports (SSTs) through the stratosphere might reduce atmospheric ozone. Subsequent investigations revealed a number of other ways that human activities might deplete ozone and thereby lead to substantial increases in rates of skin cancer and to changes in global climate. The focus of this chapter is on the scientific monitoring system that diagnosed these dangers and on the strategies that evolved to cope with them.

The subject of chapter six is a threat that poses no threat of actual harm for at least half a century-the greenhouse effect, a gradual warming of the earth's climate and the shift in

- 11 -

weather patterns caused by increased carbon dioxide and other trace gases in the atmosphere, due in part to combustion of fossil fuels. This case provides another look at the implicit strategy of using scientists to monitor technological risks and considers the real problem of how to balance risks against benefits. This chapter also raises disturbing questions about government's ability to avert global risks when these involve central aspects of our way of life.

In chapter 7 we formulate and discuss the complete catastrophe-aversion system, first describing the major elements of the system, their variations, and how each was applied in the toxic chemicals, nuclear power, recombinant DNA, ozone, and greenhouse effect cases. We then consider how our findings can be used to improve decision making about both risky technologies and other matters of policy.

In the concluding chapter we discuss prospects for improved application of the catastrophe-aversion system. One major complaint about contemporary risk management concerns the problem "How safe is safe enough?" Implementation of safety precautions is a matter of degree, and each additional precaution typically involves increased cost. One prominent school of thought holds that sensible decisions about "How safe?" can best be made by improving risk analysis. We explain why this is a dubious hope, and we propose a more strategic approach to improving the catastrophe-aversion system.

Caveats

Before turning to the subject of toxic chemicals in chapter 2, some clarification about the scope of our analysis is in order. Even though we consider a fairly wide array of risks, we make no claim to comprehensiveness. This volume analyzes selected aspects of the regulatory histories of selected technologies. There are too many potential technological catastrophes to include them all, and each of our cases is complex enough to justify its own separate work. But an extended analysis of the five technological risks reviewed here should prove useful in spite of the necessary selectivity.

- 12 -

This volume is also selective in that it focuses on the United States. There is much to be learned about other nations' responses to technologies with a potential for catastrophe, both for their own sake and to illuminate the strengths and weaknesses of U.S. responses. But undertaking such comparisons would have required a much different study than we have undertaken.

We do not attempt to cover the politics-interest group maneuvering, partisan lineups in Congress, the role of the media, struggles within regulatory agencies, details of litigation, or changes in public opinion-of each controversy. However fascinating, these are short-term phenomena, and our aim is to uncover deeper patterns, namely, regulatory strategies that endure.

The focus of this volume is restricted to civilian technologies that could result in unintentional harm. We do not include casualties and other damages that humans inflict on one another through military technologies. The potential for military-induced catastrophes almost surely is greater than for civilian technologies, and there is much to be learned about how to prevent such tragedies; this is not our topic here.

Also, our subject is severe threats. We take it for granted that some people will be harmed by almost every human activity (coins, for example, are dangerous household items because toddlers can swallow them). While we use no precise measure to distinguish severe threats from milder ones, we are concerned with those threats that potentially involve thousands of deaths annually rather than tens or hundreds. That is not to say that lesser risks are unimportant or acceptable, but they are not in the scope of this volume.

Our analysis is restricted to physical threats to the environment and to human health and safety. There are other threats raised by contemporary civilian technologies, some of which may be even more important to human well-being than physical health and safety. Some unions, for example, are concerned that widespread introduction of industrial robots could lead to significant unemployment, at least for the short term, among certain classes of workers. Likewise, some modern biomedical procedures are perceived by many people to violate

- 13 -

deeply held values. Such issues are important and deserve more attention, but they are not part of this inquiry.

Finally, this volume aims to be as unpoliticized as possible. While we have (sometimes opposing) political opinions on the subjects discussed, we have striven to keep these out of our analysis. We desire this book to be equally useful to pro- and anti-nuclear readers, to pro-growth advocates as well as to ardent environmentalists, to conservatives and to liberals. We are not attempting to change anyone's mind about any particular political controversy. Rather our aim is to clarify the strategies that have been evolving for coping with potential environmental threats that almost everyone agrees must be addressed in some manner. To the extent that regulators and other political participants have a clear perspective of the strategies that have been used to avert catastrophes, judicious application of these strategies in future controversies becomes more likely.

This book, then, examines how our society has learned to deal with the potential for catastrophe created by modern technology. If we are to have at least modest confidence in humanity's future, we must have a reasonable hope that severe physical threats from civilian technologies can be managed without unacceptable harm to human health or the ecosystem. We must know how decisions affecting health and safety get made. Are we learning how to avert catastrophes in general? Are policy makers discovering how to learn about environmental problems? And are they learning how to use such improved abilities to actually avert potential future problems?

As a result of TSCA, manufacturers are now legally required to demonstrate the safety of new chemicals, just as they are for pharmaceuticals, food additives, and pesticides. Any negative evidence, even the sketchiest, may be legally sufficient to keep a new chemical off the market.^[27] In practice some risks are considered acceptable if the projected benefits are significant, but uncertainties, if the decision is close, tend to weigh against the side that bears the burden of proof.^[28] So TSCA makes strict regulation easier.

Mechanics of Premanufacture Notification

The premanufacture notification (PMN) system ensures that EPA will have considerable information about a chemical's molecular structure, anticipated production volume in the first few years, by-products, exposure estimates, results of toxicology testing, manufacturing methods, and disposal techniques. EPA can require that industry conduct virtually any tests considered necessary to evaluate a new chemical's safety. Moreover, TSCA grants EPA more authority than ever before to act on the basis of such information.

EPA's review process begins with a structure-activity team of scientists who assign a toxicity rating to each chemical; another group of scientists rates the degree to which individuals and the environment are likely to be exposed to the

.26 .

chemical. If exposure is not rated high and health effects and ecological concerns are all rated low, the chemical passes premanufacture screening. Otherwise the chemical moves to third, fourth, and fifth levels of consideration, with each stage involving higher levels of decision makers and subjecting the chemical to increasing scrutiny.

Initially, the total number of PMN notices submitted was far below the expected amount. This seemed to indicate that industry was launching fewer new chemicals because of the new regulatory requirements. But the number of PMN submissions increased steadily in the early 1980s and leveled off in the range of 1,200 to 1,300 per year.^[29]

Surprisingly, only about three chemicals out of every ten processed through the PMN system have entered commercial production.^[30] According to EPA staff, the chemical companies "invest" in the PMN statement as a stage in research and development, that is, well before a decision has been made to market a chemical. "As soon as prospects for marketing loom on the horizon, they get the PMN in so that marketing will not be held up if the company does decide to go ahead with it."^[31] (Some of the submitted PMNs may yet come to market and thus increase the current rate.)

How carefully PMNs are reviewed depends partly on the amount of staff time available for the task. By 1985 there were an equivalent of 125 professional staff and 14 support staff assigned to full-time work on the PMN system. This represented an increase of approximately 21 percent over professional staffing levels of fiscal 1981 and a decrease of about 14 percent in support staff. Meanwhile, expenditures on the PMN

.27 .

program declined approximately 15 percent in real dollars between 1981 and 1983.^[32] While budget allocations and staffing levels changed moderately, PMN submissions increased substantially. At the 1984 submission rate of 1,250 PMN notices per year, just over one work month per staff member can be devoted to each new PMN chemical.

While the PMN program has fared better than other programs at EPA in budget battles (and no doubt efficiency has improved since the program went into full operation in 1981), it is questionable whether the current budget is adequate for the existing workload. The fact that only three chemicals out of every ten processed by EPA enter commercial production exacerbates the problem. In effect, scarce EPA time and talent are being "wasted" on chemicals that companies never bring to market.

Like most laws, TSCA is changing during implementation. The legislation explicitly provided authority for EPA to waive PMN requirements for certain classes of chemicals that are deemed to pose acceptably low risk and, as a result, numerous requests for exemption have been submitted. The one that would cover the most chemicals came from the Chemical Manufacturers Association in May 1981. It sought exemptions for high molecular weight polymers, low-volume chemicals of all kinds, and chemicals that are used only as production intermediates and that remain entirely on the premises of a chemical factory. The Dyes Environmental Toxicology Organization made a similar request, and also asked that EPA shorten the review period for various dyes and dye intermediates.^[33]

EPA has granted the bulk of the requested exemptions. Even though manufacturers of an exempted chemical still must notify EPA, there will be less paperwork, and manufacturing can commence at any time, as long as notice is filed fourteen days prior to actual marketing. There are significant exclusions and restrictions in the exemption process that are still unsatisfactory to manufacturers, however.

Exemptions to PMNs are obviously advantageous to the chemical industry. But given the large number of new chemicals and the even larger number of PMNs, exempting certain chemicals may be a sensible way to adjust regulatory strategy

.28 .

in that it may help concentrate attention on the more dangerous chemicals. Scientists consider high molecular weight polymers to be relatively nontoxic, and EPA is following the weight of scientific judgment in exempting them from PMN scrutiny. Exempting low-volume chemicals and site-limited intermediates represents a regulatory judgment that the costs of review outweigh the risks of no review. However, in the case of exemptions, only experience can tell whether it is a good idea; but it is a sensible trial.

Evaluating the PMN System

Evaluation of the PMN system is impeded by the degree of expertise necessary to judge the scientific quality of EPA's decisions. Evaluation is even further complicated by the very high percentage of PMN submissions that omit significant information because manufacturers claim confidentiality. Approximately 50 percent of PMNs contain at least one claim of confidentiality on chemical formula, name of manufacturer, intended uses, tests performed, amounts to be manufactured, or other information, and some PMNs claim that everything about the new chemical is confidential. The General Accounting Office and the Office of Technology Assessment-both exempt from the confidentiality restrictions-have begun to study the implementation of TSCA, but their reports cannot divulge any confidential information on which their conclusions may have been based.^[34]

It is clear, however, that the new system already has deterred production of some new chemicals. For instance, one manufacturer withdrew a PMN notice in April 1980 and did not manufacture six new plasticizers because EPA ordered a delay on production. The agency had required the manufacturer to develop and supply additional data on the chemicals' dangers.^[35] But some industry toxicologists question whether these plasticizers were more dangerous than those already on the market.

Detailed review and regulatory action against new chemicals have been relatively rare as a percentage of PMN submissions. Only eighteen (3 percent) of PMN submissions received

.29 .

detailed reviews in 1981; the number increased in 1982 to fifty (6.25 percent). The Office of Toxic Substances initiated eleven "unusual actions" during 1981 and thirty-one during 1982. These included: (1) suspensions of the review period to allow more time for scrutiny, (2) voluntary agreements under which manufacturers agreed to restrict the use of their new chemical in some way that EPA found sufficient to remove it from the category of unreasonable risk, and (3) formal rule-making proceedings to block manufacture of proposed new chemicals. In addition, six PMN notices were withdrawn by manufacturers in 1981 and sixteen in 1982; some of these would have been subject to enforcement action had they continued through the detailed review process.

The number of PMNs held beyond ninety days increased during 1983 and 1984. By early 1985 more than 10 percent of PMNs were being temporarily delayed. Whether this actually is a result of deeper scrutiny or is merely indicative of a backlog of work within EPA is difficult to discern. Still, only a very small number (less than 0.4 percent) have been rejected entirely on the grounds that the chemical presents an unreasonable risk. There are several possible interpretations: (1) manufacturers may be voluntarily refraining from production of the more risky new chemicals, at least in part because they expect that the substances would not be approved, (2) the original estimates that 5 to 20 percent of new chemicals would be dangerous were inaccurate, or (3) the PMN system is not screening out some of the riskier substances.

Strategies for Chemicals Already in Use

The task of monitoring some three hundred to four hundred new chemicals each year is difficult enough. But what of the sixty thousand or more existing chemicals, of which unknown thousands may have negative effects on human health or on the ecosystem. This task is staggering, and since attention can be devoted to only a relatively small number of chemicals each year, priorities must somehow be set. One way of setting priorities is by trial and error: wait for the conse-

.30 .

quences to become known and then deal with those that emerge soonest and are most severe. This strategy still is being used in Japan, Germany, and most other nations, and, as we saw in the case of pesticides, trial and error can be a viable way of setting regulatory priorities. However, TSCA attempts to improve on the results that could be achieved through such trial-and-error by imposing a priority-setting process.

TSCA established the Interagency Testing Committee (ITC) "to make recommendations to the Administrator respecting the chemical substances and mixtures to which the [EPA] Administrator should give priority consideration."^[36] The committee is instructed to consider "all relevant factors," including:

Production volumes;
Quantities likely to enter the environment;
Number of individuals who will be exposed;
Similarity in structure to other dangerous chemicals.

TSCA limits the total number of chemical substances and mixtures on the list at any one time to a maximum of fifty, and gives EPA just one year to respond to each ITC recommendation. Clearly, the intent is to identify and force action on high-priority testing needs and to keep EPA from being overwhelmed by the sheer size of the evaluation task.

Structure and Mechanics of the ITC

The ITC is composed of eight formal representatives and six liaison representatives from a total of fourteen federal agencies, departments, and programs. The ITC has the equivalent of a staff of about eighteen professionals, most of whom are from outside consulting organizations. The committee's budget of about $400,000 remained constant in the early 1980s as did its workload. The ITC meets once every two weeks for a full day, and most members spend additional time preparing for such meetings. But all members have heavy responsibilities in their regular agencies, so their ITC work is a

.31 .

part-time activity. These conditions are not ideal for such demanding work.

By 1986 the Interagency Testing Committee had issued eighteen semi-annual reports, naming over one hundred individual chemical substances or classes of chemicals for priority testing. To arrive at these recommendations, the first step is a computer search of scientific articles on toxicity, from which is developed a working list of several thousand potentially dangerous chemicals. These chemicals then are scored on the basis of production volume and the other criteria listed above. The highest scoring chemicals are subjected to detailed staff review, and the ITC reaches its decisions on the basis of a ten- to fifty-page dossier on each of approximately sixty chemicals per year. The ITC recommends for priority testing those chemicals that combine high exposures with probable high toxicity. In this process, nearly four thousand chemicals were considered by 1986, of which approximately five hundred were reviewed in detail.

Problems to Date

Testing Classes of Chemicals

Many of the ITC's early test recommendations were for broad classes of chemicals. Because there are so many chemicals that can pose dangers, the committee hoped to speed up the testing process by focusing on classes of chemicals rather than on individual chemicals. But such testing requires that appropriate groupings of chemicals be identified, and this is nearly impossible. When EPA began investigating how to pursue the ITC's recommendation on benzidine-based dyes, for example, there proved to be some five hundred of these dyes that were combined and marketed under a total of twenty-five thousand different trade names. A single category simply could not encompass these chemicals' diverse exposure expectations, production volumes, structure-activity relationships, and other characteristics relevant to testing. A similar problem arose with priority testing of the organic metallic compounds known as alkyltins,^[37] and, as a result, the ITC's recent testing recommendations have generally been for individual chemicals.

How Many People Are Exposed?

One of the main criteria in setting testing priorities is the number of people likely to be exposed to a chemical. No matter how toxic, a chemical that is manufactured in small quantities and contained will not create many problems. Unfortunately, however, available information about exposure levels is minimal.

The only nearly comprehensive data base available in the mid-1980s is based on a 1972 survey of five thousand workplaces by the National Occupational Health Survey (NOHS). It relied party on an indirect measurement method that now seems questionable. For example, because many degreasing solvents contain chlorobenzene, all employees in workplaces that used such solvents were assumed to have been exposed to this chemical. This assumption yielded estimates that are now considered by EPA to have exaggerated exposures by up to 1,000 percent.

The NOHS does not take into account chemical exposures outside the workplace, yet there is no other source of such information. Nor is there, for most chemicals, standard scientific literature on exposures. A Chemicals Inventory kept by EPA contains information on more than fifty thousand chemicals, but it is not updated to reflect current production or imports, and it was never intended as a means of calculating probable exposures. As a result, analysis of exposures is the "weakest part of our analysis-across the board" according to one of the EPA officials responsible for making decisions about priority testing.

EPA's Backlog of Cases

EPA is required by TSCA to respond to the ITC's priority recommendations within twelve months of the date they are added to the list, but EPA has not always met this schedule. As of mid-1980 EPA had proposed responses to only four of the thirty-three chemicals whose one-year deadline had expired. The Natural Resources Defense Council, a prominent environmental group, brought suit against EPA in an attempt to remedy the delays, and the court ordered EPA to develop a plan for timely testing.^[38] EPA complied with the order, and was fully caught up on its cases by late 1983.

Voluntary Agreements with Industry

EPA in 1980–81 decided to negotiate, rather than order, testing; the agency claims that it can get industry to test more quickly by this approach. Most ITC members find the arrangement acceptable, as does a study by the General Accounting Office.^[39] But the Natural Resources Defense Council contends that voluntary testing is a violation of TSCA and that it weakens public protection.^[40] If voluntary testing continues to work satisfactorily, approximately half of the ITC recommendations to date will have led to earlier or more in-depth testing of chemicals than would have occurred without such a priority-setting strategy.

Progress and Continuing Problems

A comprehensive analysis of U.S. policy toward toxic chemicals would necessarily examine many more issues than have been discussed in this chapter, but several conclusions are evident.

There has been a great deal of improvement in particular facets of the regulation and use of toxic chemicals as a result of trial-and-error learning. For example, Paris Green, which was once a severe threat to human health, is no longer used on fruits and vegetables. Similarly, use has been curtailed of DDT and many other persistent pesticides; current insecticides degrade into relatively nontoxic components much faster than those used in 1970, 1950, or even 1920.

Also, significant adjustments to the regulatory system are being made that should improve on trial and error. As a result of the premanufacture notification system, some offending chemicals will be screened out prior to introduction. The Interagency Testing Committee is gradually developing priority-setting procedures that should help direct governmental attention to the more dangerous chemicals and curtail use of such chemicals before problems actually arise. While TSCA is administered more laxly than environmentalists consider warranted, still, it is likely that many risky chemicals of the future will be spotted early instead of decades after their distribution throughout the economy and ecosystem.

.34 .

Legal and institutional innovations have improved the ability of federal, state, and local governments to cope with toxic chemical problems, and it is a positive development that, within the past two decades, environmental protection agencies, major environmental statutes, and environmental groups have come into existence in most industrial nations.

However, these optimistic conclusions must be tempered by four qualifications. First, since there is a twenty- to thirty-year delay between exposure to carcinogens and manifestation of cancer, we have not yet witnessed the results of chemicals used during the past several decades. Moreover, there were approximately ten times as many synthetic chemicals produced between 1965 and 1985 as in all previous human history. While the trends do not indicate an imminent cancer epidemic, we must wait for more time to pass before assessing the toll on human health.

Another qualification concerns priority setting. The effort made to set priorities is noteworthy-a genuine breakthrough in government's approach to regulation. To date, however, success has been limited. While responsible agencies are becoming more proficient at the task of setting priorities, it is too early to tell whether the results of these efforts will be significant.

Third, we have emphasized repeatedly that a central part of trial-and-error learning is the recognition of negative feedback. However, the PMN system is partially insulated from such feedback because of the confidentiality guaranteed to chemical manufacturers. This may be a predicament with no satisfactory resolution. Forcing manufacturers to reveal trade secrets would reduce their incentive to innovate and would increase their incentive to circumvent the system.

Finally, considerable-perhaps excessive-faith in science was displayed by Congress and by environmentalists who argued for premarket screening of new chemicals. The ability to make intelligent advance judgments about a new chemical depends partly on the results of tests for toxicity. But just as important are how a chemical will be used and the quantities in which it will be manufactured. PMN notices give the original manufacturer's estimate on these matters, but the uses to which a chemical is put can change. So the assignment given

.35 .

EPA is as much a requirement for guesswork on a new chemical's commercial future as it is for scientific testing of the chemical's dangers. Fortunately, TSCA established another regulatory process to monitor chemicals that are being put to significant new uses; but that regulatory process is even less proven than the PMN system.^[41]

Our conclusion is that, overall, scientific analysis has not entirely replaced trial and error in the regulation of toxic chemicals.

The controversy spread throughout the country, though rarely with the intensity reached in Cambridge. By 1977 the risks of rDNA research had been debated at the local level in Ann Arbor, Bloomington, Berkeley, Madison, and San Diego and at the state level in New York, California, and New Jersey. The debates resulted in extensive media coverage, much of it sensational: "Science That Frightens Scientists" (Atlantic ); "Creating New Forms of Life Blessing or Curse?" (U.S. News and World Report ); "New Strains of Life or Death?" (New York Times Magazine ). With the media interest came a flurry of books: Recombinant DNA, The Untold Story; Biohazard; Playing God .^[5] During 1977 alone sixteen bills for regulating rDNA research were proposed in Congress and twenty-five hearings were held. The primary bills that finally emerged in both the House and Senate called for extensive federal regulations. Senator Kennedy's bill, which was passed by the Senate Committee on Human Resources, proposed a national regulatory commission empowered to promulgate safety regulations, license and inspect facilities, and fine violators.

Kennedy's proposed national regulatory commission was reminiscent of the Nuclear Regulatory Commission; by 1977 the controversy had all the earmarks of the nuclear power debate aroused citizenry, media sensationalism, congressional concern, and vocal opposition by dissident scientists and environmental groups. Nowhere was the intensity of the controversy and its similarity to the nuclear debate more striking than at a National Academy of Sciences meeting held in Wash-

61 

ington D.C. in March 1977. The opening session was dominated by protesters. In front of TV cameras, they chanted, waved banners, demanded that the conference be "opened up to the people," and asserted:

This is just the first protest. . . . We are just the little ruffling wind before the storm of public outrage. . . . This is the most important social issue of the coming next decade. . . . We are not going to go quietly. We have means at our command to resist the change in the human species. We will not go gentle [sic ] into the brave new world, that new order of the ages that is being offered to us here.^[6]

But the promised "storm of public outrage" and the new regulatory regime never materialized. Recombinant DNA research, unlike nuclear power, quickly receded as an issue of contention. In Congress not one of the sixteen bills on rDNA research proposed in 1977 ever reached the floor. Only two bills were proposed in 1978, none the following year, and one in 1980. By 1981 there was almost no congressional interest in regulating rDNA research.^[7] Any remaining interest focused not on the physical risks but on the ethical and legal ramifications of the research and on the genetically-engineered products that would eventually be commercially marketed.

At the local political level, the flurry of regulations that seemed so imminent in 1977 also failed to materialize. The few laws that were imposed essentially required conformance to the NIH guidelines. No funds were appropriated and no local governmental agencies were created to regulate rDNA research. Meanwhile, the NIH revised its guidelines in 1978, 1980, 1981, and 1982, each time making more rDNA experiments permissible at ever lower levels of containment.

Why Did Concern Evaporate?

Why is the case of regulating rDNA research so dramatically different from that of nuclear power? One possibility is that there has been a growing conservative, antigovernment, antiregulation mood in the nation since the early 1980s. How-

62 

ever, this reason does not explain the continuing public concern over nuclear power, acid rain, and other technology-related issues. Another possible explanation is that the risks of recombinant DNA research are lower than those of nuclear power. This assumption runs counter to the opinions of esteemed scientists who have been arguing for years that the risks of nuclear power are far lower than other widely accepted activities such as burning coal for fuel, flying in airplanes, and driving automobiles. However, the public has not accepted this testimony from scientists. Why, in the case of rDNA research, should scientists' views have been accepted?

Perhaps the most striking difference between the two cases is that the scientific community reached more of a consensus about the risks of rDNA research than they did about the risks of nuclear power. Continuing disputes among the experts seem to lead to ongoing conflicts about policy. In all the cases cited in this book, scientists triggered the initial concern and then played prominent roles in the subsequent debates. Members of Congress, interest groups, and the media inevitably joined in, but because of the technical nature of the issues, they relied on the opinions of the experts. This pattern emerges repeatedly. On issues ranging from the arms race to carcinogens in the workplace to the prospects of future energy technologies, Senator X cites the testimony of an esteemed scientist from a prestigious scientific body, only to be countered by Senator Y, who cites opposite testimony by an equally esteemed scientist from an equally prestigious university.^[8]

For nuclear power, the most vivid example of such conflict among experts was the dispute over the Rasmussen report. This major study of reactor safety, sponsored in 1975 by the Atomic Energy Commission and directed by MIT Professor Norman Rasmussen, estimated that the worst plausible reactor accident would lead to thirty-three hundred early fatalities, forty-five thousand early illnesses, and fifteen hundred latent cancer fatalities. However, it gave the probability of such a severe accident as one in a billion (per reactor per year).

Nuclear power advocates seized on this report as support for their cause. The exceedingly low probability of an accident, they argued, demonstrated the safety of nuclear power. But

63 

others, including the American Physical Society, countered that because of substantial uncertainties (such as those discussed in chapter 3), the likelihood and effects of a serious accident could be considerably greater than estimated. The effects of low-level radiation alone would add several thousand to the estimated latent cancer fatalities.^[9]

By 1985 the debate over the Rasmussen report still had not been resolved. Studies by the American Nuclear Society and the nuclear industry, motivated in part by the Three Mile Island accident, indicated that the report had overestimated by as much as a thousand times the amount of radioactive fission products actually released in an accident. If agreement could be reached on the validity of the new figures, a significant easing of restrictions on nuclear power might be justified. But the report's optimistic findings were at least partially contradicted by an American Physical Society study group that found that in some types of accidents the release might be greater than had been estimated.^[10]

Initially, the rDNA research debate also was characterized by disagreements within the scientific community. It was a group of scientists who first raised the alarm about rDNA research and who insisted, despite the skepticism of some of their colleagues, on assessing the risks. But once the NIH guidelines were established, the politics of the rDNA debate began to change. By 1977 the scientific community seemed to have closed ranks on the issue, and it presented to Congress and the media a far more united front on the risks of rDNA research than it ever had on the risks of any other major technological policy issue. In one forum after another, one prestigious spokesperson after another including some who had initially urged caution argued that the earlier concerns had been greatly overstated.^[11]

To be sure, skeptics were still to be found, but many were associated with public interest groups rather than mainstream scientific organizations and were vastly outweighed in prestige and number by proponents of rDNA research. There was no equivalent in the rDNA debate to the dissenting opinion posed by the American Physical Society to the Rasmussen report's estimates of reactor safety. Instead, the majority of scientists

64 

favored expanded rDNA research, and, faced with this predominant opinion, most of those advocating extensive regulation backed down. Senator Kennedy withdrew support from his own bill to regulate rDNA research, which had been the leading piece of such legislation in the Senate.

It is clear that a consensus in the scientific community on rDNA research and a lack of consensus about nuclear power at least partly accounts for the differences in policy directions between the two technologies. But why were scientists able to come to near consensus about rDNA research and not about nuclear power? We return again to our original question: what is it about the rDNA research controversy that has made it so different from our other cases?

Several critics have suggested that the scientific community closed ranks because it had a vested interest in ending the regulatory debate. Unlike nuclear power, where regulations restrict industry, regulations of rDNA research would have restricted scientific endeavors. Regulations would have constrained both researchers working in universities and the growing number of scientists who were finding lucrative opportunities in the nascent rDNA industry. As long as they did not fear outside control, the critics contended, scientists expressed their doubts freely; once the threat of regulation beyond their control developed, scientists quickly closed ranks and covered over their differences about the risks.^[12]

Similarly, it can also be argued that because the NIH was both funding rDNA research and regulating it, this may have created a bias that led scientists to understate the risks. Scientists seeking research grants obviously had an incentive to avoid antagonizing the NIH, their potential sponsor. In 1978 Joseph Califano, the secretary of the Department of Health, Education, and Welfare, ruled that representatives of public interest groups should be added to the Recombinant Advisory Committee, but by that point the controversy was already waning. On the other hand, scientists can advance their careers by challenging mainstream views with solid research, so there are also incentives to speak out as well as acquiesce. And in the case of nuclear power, the Atomic Energy Commission was in the same position as the NIH funding as well

65 

as regulating nuclear research and development and still vigorous disputes about nuclear power risks occurred. While it would be surprising if self-interest did not have some role in the resolution of the rDNA research issue, it is not a sufficient explanation, and we must explore further to understand the rDNA research story.

Containing Hazards and Verifying Risks

The most distinctive characteristic of rDNA research is that the risks proved relatively manageable. The consequences of potential accidents could be limited, and estimates of risk could be verified. Therefore, the inherent nature of the problem was different from that posed by nuclear power or by most other risky technologies, and a consensus about the risks was much easier to achieve.

Strategies to Limit Hazards

While it is impossible to be absolutely certain that the consequences of a serious rDNA research accident will be contained, the probability of containing such accidents within acceptable limits is substantially higher than for nuclear reactor accidents. In addition to requiring that steps be taken to prevent accidents, the NIH relied on two tactics for limiting such hazards.^[13] First, and reminiscent of the early approach to reactor safety, the NIH required both physical and biological containment. The NIH classified all permissible rDNA experiments according to their level of potential hazard. The greater the potential hazard, the more extensive the containment requirements. For example, according to the original NIH guidelines, experiments in the high potential hazard class could only be performed in labs of monolithic construction that were equipped with air locks, systems for decontaminating air, autoclaves, shower rooms, and other physical containment safeguards.

Biological containment also was required for more dangerous experiments.^[14] Scientists were required to use as host

66 

organisms only bacteria that had been shown in animal and human tests to be so enfeebled that they could not survive outside the lab. In this way, even should an accident occur during a hazardous experiment and physical containment fail, biological containment (the inability of the organism to survive outside the lab) would limit the hazard. Even experiments in the least hazardous category had to use only approved bacteria as the host organism. The most common host organism was E. coli K-12, a laboratory strain of a common, well-studied colon bacterium (Escherichia coli ).

The NIH also limited potential hazards by prohibiting particularly risky experiments. Originally, six classes of experiments were banned: (1) experiments with more than ten liters of culture; (2) experiments in which organisms containing rDNA were deliberately released into the environment (such as the one delayed by Judge Sirica in 1984); (3) experiments using DNA from certain pathogens; (4) experiments using DNA segments that code for vertebrate toxins; (5) experiments using rDNA techniques to create certain plant pathogens; and (6) experiments in which drug resistance traits were transferred to disease-causing organisms. These prohibitions were not absolute, since the NIH would make exceptions if sufficient safety precautions were demonstrated.

Prohibiting risky experiments was a very simple method of limiting hazards. To apply this method to nuclear power, reactors would either have to be limited to sizes that would virtually guarantee containment or be prohibited from any place except very remote areas.

Being able to limit the consequences of an accident from rDNA research simplified the NIH's task. The alternative would have been to rely primarily on prevention, which, as we have seen for nuclear reactors, is a difficult strategy to implement in practice. Limiting harm from accidents is simpler than trying to prevent accidents altogether. Regulators were thus able to focus on the relatively limited problem of estimating and protecting against the worst potential consequences of accidents. This is by no means a trivial task, but it is more manageable than trying to anticipate all possible causes of accidents.

Ways to Verify Risks

For both nuclear reactors and rDNA research, reputable experts argued that the risks were acceptably low, but such claims were much more easily verified for rDNA research than they were for nuclear power.

Substantial efforts have been made to assess the risks of rDNA research. At least sixteen categories of risks have been identified as shown below:^[15]

Gordon Conference, 1973. Risks: new types of plasmids and viruses; large-scale preparation of animal viruses.

Asilomar Conference, February 1975. Risks: spread of antibiotic-resistant organisms; alteration of the host range of bacteria; and "shotgun" experiments with unknown outcomes.

Senate Health Subcommittee, April 1975. Risks: infections not susceptible to known therapies; animal tumor virus genes in bacteria.

Director's Advisory Committee Meeting, February 1976. Risks: new routes of infection; novel pathogens; disease transmitted by E. coli.

Senate Health Subcommittee, 1976. Risks: spread of "experimental cancer"; virulent hybrids due to combining two mild viruses; unknown impact on biosphere if new species created.

House Science and Technology Subcommittee, 1977. Risks: altering the course of evolution; transmission of laboratory recombinants to wild strains of bacteria; colonizability of E. coli.

National Academy of Sciences Forum, 1977. Risks: latent tumor viruses; insulin genes in E. coli; extraintestinal E. coli infections; breakdown products of recombinant DNA molecules.

Senate Health Subcommittee, 1977. Risks: possible interference with human autoimmune system; unanticipated hazards; disturbance of metabolic pathways.

68 

Falmouth Workshop, 1977. Risks: creation of more dangerous bacteria; spread of R-factors.

Ascot Workshop, 1978. Risks: penetration into intestinal lining; new types of experiments with viruses.

Workshop on Agricultural Pathogens, 1978. Risks: E. coli transformed into a plant pathogen; more virulent strains of plant pathogens.

These hazards have been discussed at more than a dozen scientific workshops and conferences, analyzed through experiments, and reviewed by the Recombinant Advisory Committee, the NIH, the Office of Technology Assessment, and other federal organizations. Much of the research has attempted to answer one or more of the following questions:

Could an organism escape from a laboratory and establish itself in humans, animals, or other parts of the natural environment?

Could a recombinant organism transfer its rDNA to other organisms?

Could the rDNA make the escaped organism dangerous to man or the environment?

From the outset, some scientists asserted that the answer to all of these questions is "no." But the majority of scientists originally said, in effect, "We do not know, so we will have to proceed cautiously until we find out." Their attempts to answer these key questions are an important part of the rDNA story and reveal a great deal about the task of averting catastrophe.

Survival of Organisms Outside the Lab

The most likely way that organisms could escape from a research laboratory, many scientists believed, was by their unwitting ingestion by a lab worker. At least four sets of studies were performed in which large amounts of E. coli K-12 (up to ten billion organisms) were fed to volunteers. All of the organisms died in a few days, and none were passed out in the stool of the volun-

69 

teers. In another study, fecal cultures from sixty-four lab personnel working with E. coli K-12 were periodically obtained over two years. The workers used no precautions other than standard microbiological techniques. At no time in the two years was E. coli K-12 recovered in the stool, and this again suggested inability of the bacteria to survive outside the lab.^[16] These findings, among others, were interpreted as follows in a letter to the NIH from the chairman of a conference of infectious disease experts held at Falmouth, Massachusetts, in June 1977: "On the basis of extensive tests already completed, it appears that E. coli K-12 does not implant in the intestinal tract of man."^[17]

There were some qualifications to this conclusion, however, as was attested to in the chairman's summary of the published professional proceedings of the Falmouth conference:

A number of variables are known to influence the colonization of organisms in the intestinal tract. Implantation can be altered by antibiotic administration, starvation, the type of diet, reduction in gastric acid, and antimotility drugs. It is clear that more implantation experiments need to be performed with attention to these variables, many of which may be found in laboratory workers exposed to these organisms.^[18]

The report acknowledged that certain strains of E. coli can implant under certain conditions. Persons undergoing antibiotic treatment, for example, are susceptible.^[19] The NIH subsequently established laboratory practice guidelines to protect against some such dangers, but it is questionable whether the guidelines are enforceable. Overall, however, a great majority of scientists working in this field concluded that the probability of hazard was too low to take precautions beyond those that are standard in microbiological work.

Transmission to Other Organisms

Even if some of the organisms escaped, they would soon die unless they could establish themselves in a human, animal, or other part of the natural environment. Research has also been conducted to evaluate this possibility. In more than thirty years of use in genetics laboratories, E. coli K-12 has a perfect safety record,

70 

and in a study of almost four thousand laboratory-acquired infections, only two involved any type of E. coli. In neither of these cases was the infection passed on to another person.^[20] Moreover, large numbers of organisms would have to be ingested in order for a human to develop an intestinal illness. This requires more than personal contact usually a grossly contaminated source of food or water. Given contemporary sanitation and food standards, it was judged highly improbable that an infected laboratory worker could start an epidemic of gastrointestinal illness.

Again, however, this conclusion must be qualified. E. coli can cause diseases outside the intestine, and in these cases it takes fewer organisms to produce such illnesses. But most of the rDNA research scientists concluded that transmission of genetically altered organisms would be very unlikely.^[21]

Creating a Dangerous Organism

Even if escape and transmission of organisms containing recombined genes is very unlikely, it is not impossible. If a dangerous organism were created and somehow did escape, the consequences conceivably could be catastrophic. Therefore, research was conducted to examine the likelihood that a dangerous organism might unwittingly be created. In one class of rDNA experiments, known as "shot-gun" experiments, the genes of a plant or animal cell were cut into many segments, and these were inserted at the same time randomly into E. coli K-12. These experiments in particular raised fears of an unwitting creation of harmful organisms. To test these fears, the Recombinant Advisory Committee recommended a so-called worst-case experiment that the NIH agreed to fund.

The intention of the worst-case experiment was to see if a dangerous organism could be created deliberately. Investigators inserted the DNA for a virus into E. coli, and then administered the altered organisms to mice. Only one of thirteen experimental combinations produced illness and at a rate much lower than the virus would have caused without the DNA manipulation. Most scientists found the experiment reassuring; as one biochemist put it: "This type of experiment is therefore safer than handling the virus itself . . . and a major

71 

lowering of the required containment levels seems clearly justified."^[22] Again, however, this conclusion required qualification. According to several critics, the one case of viral infection was a new disease pathway not possible without the rDNA techniques.^[23]

In another series of experiments to evaluate risks, scientists inserted a similarly altered organism in hamsters to test tumor formation. No tumors developed under normal experimental conditions, but under less likely experimental conditions, where two copies of the virus were inserted simultaneously (as might accidentally occur) tumors were produced. The researchers interpreted this evidence as supporting the safety of the research, as did most other scientists. However, a small minority again dissented and criticized the experimenters' methodology and interpretations.^[24]

Because of studies such as these, rDNA research seemed less threatening, and subsequent research reinforced this impression. The gene structure of plants and animals was shown to be significantly different from the gene structure of bacteria. Plant or animal DNA must be altered deliberately to make it compatible with the host if it is to become an active part of the bacteria into which the genes are inserted.^[25] Similarly, in most rDNA experiments, the new segment of genes constitutes no more than 0.5 to 1 percent of the total gene set of the host organism. Unless this segment is carefully and deliberately integrated with the host's genes, the host is unlikely to be significantly affected.^[26]

Another worst-case scenario was that E. coli bacteria containing rDNA designed to produce insulin were assumed to have escaped from the lab, established themselves in the human intestine (and replaced all the normal intestinal E. coli ), and produced the maximum amount of insulin. Even given these assumptions, the analysts concluded that the total amount of insulin that could be produced would "not be expected to have much, if any effect, on a mammalian host." The results of the analysis, when generalized to more active proteins, indicated that most experiments would pose little problem.^[27]

The risk that rDNA can turn a host into a pathogen was

72 

summarized at another important meeting held in 1978 at Ascot, England, and attended by twenty-seven scientists, most of whom were virologists:

The workshop concluded that inserting part or all of the gene set of viruses into E. coli K-12, with approved vectors, poses "no more risk than work with the infectious virus or its nucleic acid and in most, if not all cases, clearly presents less risk. In fact, . . . cloning of viral DNA in E. coli K-12 may produce a unique opportunity to study with greatly reduced risks the biology of extremely pathogenic and virulent viruses." In other words, inserting a piece of viral DNA into E. coli K-12 locks it up inside the bacterium and makes it much safer to handle than the naked virus itself.^[28]

A few scientists did not share this view, but the great majority believed that the evidence showed it to be very unlikely that genetically altered bacteria would be dangerous in the unlikely event that they did escape and the even more unlikely event that they established themselves in the environment.

Appraisal of risks

These experiments and analyses do not show that there is no risk from rDNA research, and in fact they reveal that risks can arise in certain, very specific, highly unlikely conditions. But such controlled experimentation and the data generated through experience led a great majority of scientists and experts to conclude that most of the early concerns about rDNA research were unfounded. The ability to practically assess risk is the essential difference between the nuclear power and rDNA research issues. Much of the analysis of reactor safety is hypothetical, focusing on unlikely conditions that might arise during events that have never occurred. In developing policy about rDNA research, by contrast, policy makers have been able to make such analyses on a more concrete basis; risks have been evaluated by empirical tests governed by usual scientific methods and standards.

The concrete nature of this evaluation has had an enormous effect on the regulatory process. As confidence grew that E.

73 

coli K-12 was unlikely to survive outside the lab or be dangerous if it did, the guidelines governing rDNA research were gradually relaxed. In 1978 the requirements for biological containment in experiments using E. coli K-12 were reduced to the lowest level, and in 1981 most classes of experiments using E. coli K-12 were exempted from the guidelines altogether. Moreover, new host systems were approved for use, the NIH abolished or reduced restrictions on most experiments not using E. coli K-12, and by spring 1982 there was no longer a category of prohibited experiments.

Conclusions

In chapter 7 we will describe the major characteristics of a catastrophe-aversion system that is at least partially illustrated in each of our cases. The history of recombinant DNA research provides the clearest example of such a system in operation. When concerns about the risks of rDNA emerged, steps were taken to protect against potential catastrophe. Certain classes of experiments were altogether banned, and all others had to be performed within biological and physical containment. Once precautions had been taken, attempts were made to learn more about the risks, both by deliberate testing and by monitoring experience. As more was learned about the risks and as concerns decreased, the initial precautions were progressively relaxed.

Critics have argued that this strategy should have been implemented more stringently. It is true that testing could have been more thorough, the burden of proof could have been kept more squarely on the advocates of potentially risky research for a longer period, nonscientists could have been given more decision-making authority earlier in the process, more of the opposing scientific opinions could have been evaluated, more experiments could have been prohibited, and the guidelines could have been relaxed more slowly.

How cautiously to proceed in protecting against potential risks is always a matter of judgment as well as a matter of science. In the case of rDNA research, a substantial majority

74 

of experts in the field agreed that the risks were very low. The NIH, local governments, and members of Congress went along with this judgment, and these judgments have been borne out by experience. Former critics of rDNA research no longer are vocal about the dangers perhaps persuaded at last or maybe just overwhelmed. Whether or not the critics were correct, sensible strategies were developed to cope with the uncertainties and potential risks of rDNA research. We discuss these strategies in more detail in our concluding chapters.

A secondary point in this chapter has concerned the striking difference in the fates of the public debates over rDNA research and nuclear power. We have argued that this difference can be attributed largely to the differing natures of the two problems. In rDNA, a worst-case event can be contained and estimates of risk can be tested empirically. A worst-case event for a nuclear reactor cannot be contained and therefore must be prevented; and estimates of risk must necessarily remain hypothetical. Although one side or the other in the nuclear controversy may be mistaken, there is a sound reason why no strategy has been developed that can resolve the debate.

If the objective nature of the technological issue affects its fate, what about the many other explanations for technology policy debates? And how does the objective nature of a technology relate to the widely accepted view that public perceptions and fears are what really guide technology debates and shape policy? We believe that the nature of a social problem limits what constitutes legitimate debate. The nature of the problem in the nuclear power debate is such that there is no way to establish definitively the magnitude of the risks. Advocates of nuclear power can insist that the probabilities of accidents are very low, that the major causes of accidents have been anticipated, and that the worst case would not really be that bad, but none of these arguments can be fully verified. Regulators are left with no conclusive basis for deciding between these claims and opposite claims. Lacking a basis for resolving the facts of the matter, factors like public perceptions and general attitudes become important. The position one takes on the conflicting estimates of the risks depends on

75 

whether one trusts government institutions, whether one fears high technology, and so on.

In contrast, the nature of the rDNA problem imposed objective constraints on the resulting debate. Once studies demonstrated that E. coli K-12 was highly unlikely to survive outside the lab and cause epidemics, the credibility of counterclaims about high risks diminished. The burden then shifted to the opposition to provide evidence, based on experience or experiments, that the risks were in fact as great as claimed. In other words, the facts constrained the debate. Such "facts," determined at any given time by scientific consensus, may later prove to be mistaken; until that time, they restrict the range of legitimate disagreement. And they constrain the currency of the debate as well, since only new "facts" are sufficient to displace old ones.

It is important to note that the discussion in this chapter has been limited to the physical risks of recombinant DNA research. As commercial biotechnology products are introduced, a new set of potential risks and regulatory issues is emerging. Moreover, as scientific techniques become increasingly powerful, debates surely will arise over the propriety of intervention into the human gene structure.^[29] Whether appropriate strategies will be developed for handling these ethical and philosophical issues remains to be seen. To the extent feasible, the strategies used in the rDNA research controversy will be worth emulating.

The Original Strategy
Reduction of Uncertainty

The principal strategy employed against the greenhouse effect has been to attempt to reduce uncertainty about the nature, magnitude, and timing of this threat by increased scientific efforts at research and monitoring.

Early Scientific Research

The first description of the greenhouse effect was given in 1863, at almost the same time that the technologies of the Industrial Revolution began to create the greenhouse problem. Tyndall, a British scientist, was the first to recognize that water vapor transmits to the earth a substantial portion of sunlight received in the upper atmosphere, while blocking infrared radiation from the earth that otherwise would escape

101 

back into space.^[9] Because carbon dioxide molecules have the same radiation-transmitting properties as water vapor, scientists soon hypothesized that increased CO₂ would cause a warming of the earth's surface and the surrounding atmosphere because less heat could be reradiated into space. The Swedish scientist Arrhenius calculated in 1896 that a doubling of CO₂ would cause an increase of 6 °C (11 °F) in the mean annual global temperature.^[10] In 1899 Chamberlin reached the same conclusion as a result of work on glaciation. "What caused glacial periods and warming periods?" he asked. His answer was: CO₂ fluctuations.^[11]

Early in the twentieth century, several scholars perceived the importance of increasing fossil fuel combustion and suggested that dramatic increases in CO₂ might occur. Others, however, were concerned about possible diminution of atmospheric CO₂ due to a decline in volcanic activity. In 1938 Callendar, a British meteorologist, used recent measurements of global temperature changes to argue that the slight warming of the global atmosphere since the mid-1800s could be accounted for by increasing combustion of fossil fuels.^[12] However, his case won few adherents.

As part of the postwar boom in scientific research, a number of U.S. scientists began in the 1950s to investigate the carbon cycle. A scientist at Ford Motor Company made the first sophisticated calculations of surface temperature responses to increased CO₂ ; he estimated a 3.6 °C increase for doubled CO₂ .^[13] Enough research on the issue was underway by 1957 so that two scientists could confidently term CO₂ released by human activities a "large-scale geophysical experiment."^[14]

In 1963 the Conservation Foundation sponsored a conference on this subject and brought it into wider public awareness.^[15] This awareness grew when in 1965, the president's Science Advisory Committee cited the CO₂ problem as a potential threat to the Antarctic ice cap.^[16]

In 1970 the MIT-based Study of Critical Environmental Problems joined the growing number of scientific organizations that had come to subscribe to the CO₂ -warming hypothesis.^[17] Then in 1971 the Study of Man's Impact on Climate drew scientists from all over the world to survey the state of

102 

knowledge in this field of study. This group of scientists put together detailed recommendations concerning the additional research and the CO₂ monitoring efforts necessary to confirm the extent of the threat and to improve overall understanding of the CO₂ cycle and its associated problems.^[18] At about this time, the U.S. Committee for the Global Atmospheric Research Program also recommended a set of research priorities, as did the international Joint Organizing Committee of the same organization.^[19] Among many other efforts in the ensuing decade, major international workshops were held in Germany in 1976 and 1977 to update research on CO₂ .^[20]

The U.S. government has been ahead of all others in fostering research and discussion on the greenhouse threat. Reports on the topic have been issued steadily since the early 1970s by numerous federal agencies and government-affiliated scientific organizations. Research expenditures as of 1984 were estimated at $20 million annually by the Office of Science and Technology Policy.^[21] The Department of Energy has an entire division devoted to carbon dioxide research, and congressional committees have held hearings on the greenhouse issue since the mid-1970s.^[22]

A Consideration of the Uncertainties

Researchers gradually identified major areas of uncertainty about the greenhouse effect, one of which is the fundamental question: will increases in atmospheric carbon dioxide actually lead to increases in temperature? The only way to know for sure is to detect confirming evidence that the world's climate actually is beginning to get warmer. Part of the research endeavor, therefore, has been devoted to the search for a "CO₂ signal."

Global temperature rose approximately 0.6 °C (1 °F) between 1880 and 1940. In the ensuing thirty years, temperatures declined by about 0.3 °C. Then another warming trend set in about 1971, and global temperatures increased by 0.24 °C just during the 1970s. Because of the cooling trend between 1940 and 1970, a number of scientists have questioned whether the temperature increase after 1970 was due to human activities.

103 

They thought it might be due to changes in intensity of solar radiation, to dust particles from volcanic eruptions, and to other natural phenomena.

Fortunately, this type of question can be partially checked against experience. The measured increase in temperature generally corresponds with the climatic warming that contemporary scientific models calculate should have occurred over the past century due to use of fossil fuels and other greenhouse gases. For example, one group of researchers has attributed about 60 percent of the warming since 1970 to a 12 parts per million increase in atmospheric carbon dioxide. The remaining warming they believe to have been caused by other greenhouse gases.^[23] So there is modest (but far from universal) scientific agreement that a warming effect is already occurring.

The extent of future warming remains in doubt, however, partly because of the overall imprecision of climate models. There are also specific uncertainties, such as whether the warming trend will create other effects that slow down the warming. Warmer, moister air may create denser clouds, for example, thereby screening out a portion of incoming solar radiation; according to one researcher, this could "reduce the expected warming over the next century by as much as one half."^[24] There may be other feedback effects that magnify or speed up the warming trend.

Another key uncertainty concerns the changes in regional climate, especially rainfall, that would result from increases in temperature. Reconstructions of regional climate patterns using geological evidence from warmer periods partly confirm projected changes, but there still is a long way to go in refining scientific understanding of regional climate patterns. Moreover, it is not yet clear how changes in the amount of rainfall translate into changes in the amount of usable water; the common assumption is that stream runoff would diminish appreciably, but some water resource analysts predict just the opposite. Also, since more carbon dioxide in the air will enhance plant metabolism, there is a (yet to be confirmed) possibility that less water may be necessary for agriculture. However, leaves grown in laboratories under enriched carbon dioxide conditions have less protein, so insects tend to eat more, and

104 

therefore plant production "could even be reduced below current levels."^[25]

Another element of uncertainty was identified in the mid-1970s when a number of scientists suggested that the cutting down of the world's forests could be decreasing natural consumption of CO₂ and thereby increasing the rate at which CO₂ is added to the atmosphere. Insects and bacteria would also add CO₂ to the atmosphere as they consumed the decaying vegetable matter remaining from deforestation. However, more recent studies have argued that forest regrowth has ballanced cutting in recent years. While there still is some controversy on this point, most researchers believe that combustion of fossil fuels will be the only major factor in future changes in atmospheric CO₂ levels.^[26]

Still another area of uncertainty concerns the contribution of gases other than carbon dioxide to the greenhouse effect. It was not until the late 1970s that scientists directed systematic attention to the potential greenhouse effects of methane, nitrous oxide, and other trace gases. Even up to 1985 these gases had received considerably less attention than CO₂ . In spite of the NRC's recognition that controlling emissions of other trace gases may be cheaper and easier than controlling emissions of carbon dioxide, less than ten pages in their five hundred-page 1983 report were devoted to this matter.

Atmospheric methane concentrations have increased substantially over the past several centuries and could double by the year 2050. This would contribute about 0.2 °C to global warming.^[27] The rate of methane released is almost certain to go up, since much of it occurs during production of important food sources. Moreover, increased carbon mon oxide in the lower atmosphere will reduce the concentration of compounds that destroy methane, and as global warming occurs, extensive peat bogs in northern latitudes may thaw and release huge quantities of trapped methane. Also possible are methane releases from continental slope marine sediments.^[28]

The effects on global temperature of fluorocarbons are difficult to predict. In one study, these chemicals were projected to contribute to an eventual temperature increase of 0.3 °C if their production remained at 1973 levels.^[29] Because of the ozone

105 

controversy and sluggish economic growth, usage of fluorocarbons dropped until 1983, when it began to rise rapidly. Production increases of 3 to 5 percent annually are considered possible but not certain. Moreover, as described in chapter 5, the distribution of ozone at different altitudes is changing, in part because of fluorocarbons. Because the effects of ozone on surface temperature vary at different altitudes, the overall effects of fluorocarbons on climate are very complex.

By 1985 the most comprehensive study on the combined effects of all the greenhouse gases concluded that the effects of the other trace gases "are as important as that of CO₂ increase in determining the climate change of the future or the past one hundred years." This group of researchers calculated a 0.8 °C (1.4 °F) warming due to trace gases by 2030.^[30] But another prominent scientist, who chaired the National Research Council's 1983 study, countered that he "would be inclined to take [these results] with a grain of salt. You're on surer ground with CO₂ ."^[31] It is thus evident that there is not yet agreement on the scientific issues involved in the effects of trace gases and even less consensus about future rates of increase in their production and use.

Another uncertainty about the warming effect concerns the probability and extent of flooding due to melting polar ice. There is disagreement about the extent of warming in the southern hemisphere. Some researchers claim that warming in the Antarctic will approximate the global mean; if this is correct, not much melting would occur. A majority of researchers, however, expect that warming at the poles will be more than twice the global average; this could cause major melting.^[32] Moreover, current theories cannot predict whether the land-based Greenland and East Antarctic ice sheets would shrink or grow with a warmer climate. If the ocean warms while the air above the ice sheets remains below freezing, for example, increased snowfall could enlarge these icecaps, with a consequent reduction in sea level.

By the mid-1980s, then, a pluralistic process of research and monitoring by diverse groups of scientists had highlighted the (large) remaining uncertainties about the greenhouse threat. Despite substantial improvements in meteorology and other

106 

relevant sciences, however, relatively little progress was made in actually reducing the uncertainties about the timing, magnitude, and effects of the projected climate changes.

Need for More Focused Research and Monitoring

Identification of the above-mentioned uncertainties has made them, to some extent, the focus of greenhouse effect research. But since the number of potentially useful research projects on this topic is so large, some way to set priorities is necessary. Congress suggested one approach for priority setting in the Energy Security Act of 1980. Because the synthetic fuels industry (which this law sought to establish) would contribute to the greenhouse threat, Congress directed the Office of Science and Technology Policy to coordinate with the National Research Council to comprehensively study the carbon dioxide problem.^[33] Congress asked that the following issues be addressed:

A comprehensive assessment of CO₂ releases and impacts;

Advice on how to structure a long-term program of domestic and international CO₂ research and assessment, including definition of the U.S. role and the necessary financial resources;

Evaluation of "how the ongoing United States government carbon dioxide assessment program should be modified so as to be of increased utility in providing information and recommendations of the highest possible value to government policy makers."^[34]

Through a newly formed Carbon Dioxide Assessment Committee, the NRC issued a 1983 report generally considered to be a very good overview of the greenhouse threat.^[35] But the NRC fully addressed only the first of the tasks that Congress posed (namely, comprehensive assessment). No attention was devoted to the third item, and instead of analyzing the structure of an international research program (as suggested in the second item), the committee merely constructed a very long list of the research topics it considered desirable to pursue. Among some one hundred topics were the following:

107 

Long-range economic and energy simulation models for projecting CO₂ emissions;

Studies of ice cores, tree rings, and lake sediments to refine historical knowledge on past CO₂ concentrations;

More sophisticated ocean modeling, to better predict rates of CO₂ uptake and warming of various ocean layers;

Modeling and data collection on cloudiness;

Sediment sampling programs on continental slopes to learn more about vast quantities of trapped methane that could be released as oceans get warmer;

A wide variety of studies of the Antarctic ice sheet;

Effects of carbon dioxide on photosynthesis and plant growth;

Climatic effects on agricultural pests;

Possible human health risks from higher CO₂ levels;

Extensive water conservation research.

In addition to research on the carbon cycle, the NRC committee called for more extensive and more sophisticated monitoring to determine more conclusively whether climatic changes actually are beginning to occur. As one part of the NRC report put it, "Policy makers are not likely to take action unless we can demonstrate that CO₂ actually is making the climate warmer."^[36] Among the types of monitoring called for by the NRC were:

Expanded, ongoing temperature measurement to detect a CO₂ "signal" at the earliest possible date;

Systematic, ongoing oceanic measurements throughout the world from research ships, ocean-scanning satellites, and other sources;

Improved monitoring of non-CO₂ greenhouse gases.

No doubt these all are worthwhile measures to extend scientific understanding, with or without a greenhouse problem. But this leaves us with a long list of research topics and with little prospect of quickly narrowing the uncertainties. Are some uncertainties more important than others? Is there a

108 

subset of research topics that would allow policy makers to come more quickly to an understanding of the issues? For example, it is likely that some of the trace gases deserve more attention than others. It is known that some of these gases will be easier to limit than others, so it is sensible to invest scientific resources where there is more potential for control. Since each gas reflects back to earth only certain wavelengths of escaping heat radiation, moreover, some "windows" may be of greater concern than others. For example, gases that reflect on approximately the same wavelength as carbon dioxide may not result in as much additional climatic warming as gases that block other windows. If so, then both research and control strategies might focus on the especially troublesome gases.^[37]

These are the kinds of priorities that must be established and addressed before policy makers can begin to act. The failure to date to establish realistic research priorities based on the need of policy makers is a significant shortcoming in the strategy deployed against the greenhouse effect.

Alternative Intervention Strategies

If and when the uncertainties about the greenhouse effect are reduced and agreement to act against this threat is reached, what control strategies are available? At least four are already under discussion among scientists and policy analysts.

Promoting Alternatives to Risky Activities

The most obvious strategy would be to examine safer ways to accomplish those economic functions that now are creating the greenhouse effect. It is debatable whether this is a practicable approach. To pursue it would require aggressive energy conservation and substantial increases in research and development on nonfossil energy sources. Solar and biomass energy sources are usually envisioned, but greater reliance on nuclear power might also be included.

Some energy analysts believe that strategy should be initiated at once, not only to reduce the greenhouse threat, but

109 

also to reduce reliance on foreign oil and limit acid rain. Others see the conservation/alternative fuels option more as a backup strategy; as the 1983 NRC committee expressed it: "We may find that emissions are rising rapidly, that the fraction remaining airborne is high, that climate is very sensitive to CO₂ increase, or that the impacts of climate change are costly and divisive. In such a case, we want to have an enhanced ability to make a transition to nonfossil fuels."^[38]

Most energy analysts expect continuing improvements in energy efficiency and alternative energy availability, but the view held by many is that increases in demand for energy will outstrip such improvements. The NRC 1983 report, for example, estimated that world fossil fuel consumption would approximately triple by the year 2100 in the absence of major bans, taxes, or other disruptions of consumption. But the Harvard Energy Policy Project and some scenarios from the International Institute for Applied Systems Analysis are more optimistic about the potential for replacing fossil energy sources with renewable ones.^[39] A 1984 Stanford/MIT study calculated that changing patterns of energy usage from fossil fuels to other existing technologies by just 1 percent per year would reduce climatic warming greatly and extend it over a much longer period at reasonable costs.^[40]

It is worth recalling, moreover, that as recently as 1975 after the initial OPEC oil embargo most analysts overestimated the U.S. energy demand for 1985 by more than 50 percent. If this much error can occur in ten years, current projections for the years 2050 or 2100 could be even more flawed.^[41] Thus, the feasibility of this conservation/renewable energy strategy is debatable but not clearly disproven.

Limiting Dangerous Activity

If the conservation/renewable energy strategy is not sufficient, an obvious supplement to this strategy would be to limit risky activities in some way. This was the strategy followed in restricting the amount of pesticide residue on fruit, the quantity of air pollutants released from a factory, and regulations of numerous other health and safety threats. In the

110 

greenhouse effect case, this strategy could result in limiting the amount or type of fossil fuel burned each year as well as limiting use of other greenhouse gases. At least two tactics taxes and bans can be used in implementing this strategy.

As recent successes in energy conservation demonstrate, one way to reduce the use of fossil fuels is to increase their price. This will occur to some extent as a result of supply and demand for depletable resources, but prices could be increased even more by levying a tax. For example, EPA has investigated the following option that would double the cost of shale oil, a (future) fuel high in carbon:^[42]

Lower carbon fuels would be taxed in proportion to their carbon content.

How much effect would levying such taxes have on reducing use? The only way to judge this is to rely on an economic model that forecasts future energy usage via computer simulation, and no such model has a record of demonstrated reliability.^[43] However, one well-regarded model used by EPA predicts that such a tax applied worldwide would reduce CO₂ emissions by some 18 percent by the year 2050 and by 42 percent by the year 2100.^[44] This would reduce global temperature by 0.7 °C by the year 2100, if global warming is on the low end of the anticipated range; if warming is greater, the reduction would be greater. Higher taxes would lead to greater temperature reduction; a tax imposed only in the United States would have about one-third the effect of a world tax.^[45]

111 

An even more stringent approach would be to ban the use of high-carbon fuels. As a result, solid fuel prices would more than double due to scarcity and total demand for energy would drop by about 50 percent. A major shift to cultivation of plants for both solid and liquid fuels would be expected, and this form of biomass energy does not add increased carbon dioxide to the atmosphere. As a result, projected CO₂ emissions for the year 2100 could be cut about in half by a ban on coal or shale oil. If both were banned, CO₂ emissions could be reduced substantially from their present level according to EPA calculations. Banning both would delay the projected date that CO₂ would double by about twenty-five years and would cut almost in half the overall temperature increase by the year 2100.^[46]

The drawback of these particular strategies is that they would be very costly. Banning coal or heavily taxing its use would have massive negative economic consequences that perhaps would rival the impact of the greenhouse effect itself. The likelihood that such measures would be implemented even in the United States, let alone in other nations with fewer energy alternatives, is small.

Partial bans or taxes could be applied to some of the other greenhouse gases. Taxes to increase the price of such gases presumably would cut usage. Limits on authorized production levels, such as those used for fluorocarbons in Europe, might be a simpler way to achieve the same outcome. How much of an effect could be achieved at what cost is even more of an unknown for these other trace gases than for CO₂ . Yet production of some of these gases is growing rapidly, and they could have as much influence on climate as carbon dioxide. So serious attention to control strategies appears overdue.

Preventing Dangers

The strategies discussed above rely on limiting risky activities, such as fossil fuel combustion. An alternative strategy would be to allow these activities to continue but to make them safer, and some analysts believe that it may be possible to prevent fossil fuel combustion from leading to the green-

112 

house effect. For example, one suggestion in 1977 was to install "scrubbers" (equipment for removing carbon dioxide) on coal-fired power plants; the captured CO₂ would be injected deep into the ocean.^[47] The technical obstacles to this approach are probably insurmountable, but the underlying goal is worth pursuing.

One tactic would focus on long-term removal of carbon dioxide from the atmosphere. Because trees metabolyze carbon dioxide and incorporate the carbon into their fibers, it is conceivable that enough trees could be planted to significantly reduce the amount of CO₂ in the atmosphere. The idea of using forests as a sink for CO₂ emerged in the mid-1970s when concerns arose about the effects of deforestation.^[48] Such a reforestation effort would build up the quality of soils and help prevent erosion besides being aesthetically appealing. Moreover, as a decentralized solution to a global problem, it is politically attractive.

Unfortunately, the obstacles to this approach are severe. If American sycamores were used because they grow well in temperate climates with minimal rainfall, one analysis estimated that a land area roughly the size of Europe would be required to offset fifty years of carbon dioxide emissions (at current, not increased, rates).^[49] Enormous quantities of fertilizer would be required. Acreage requirements would be reduced substantially by using a tropical tree such as the Hawaiian leucaena, which absorbs four times as much carbon, but irrigation and fertilizer requirements would increase substantially. A 1983 EPA estimate put the initial cost as high as $400 billion, with annual expenses perhaps reaching $100 billion.^[50]

As large as this sum is, it is a small percentage of worldwide energy expenses, and there would be offsetting revenues from harvesting such forests. A recent EPA study nevertheless concluded that "sequestering atmospheric CO₂ by trees is an extremely expensive, essentially infeasible option for controlling CO₂ ."^[51] While obviously expensive, reforestation could actually prove to be a bargain in comparison with other alternatives such as banning the use of coal.

Another approach to preventing the greenhouse threat (at this stage no more than speculation) would be to actively

113 

intervene in the atmosphere to offset the warming trend. Several scientists have suggested injecting sulfur dioxide into the stratosphere to reflect a portion of incoming sunlight, but costs and environmental effects (such as acid rain) would have to be carefully examined.^[52] Alternatively, temporary cooling results naturally from the enormous quantities of dust spewed into the atmosphere by major volcanic eruptions. Would deliberate injection of dust particles into the atmosphere be technically feasible? If so, would it be relatively benign environmentally? There has not been extended study of such corrective options, and, while all such possibilities might prove infeasible or even highly dangerous, they deserve careful scrutiny.

Mitigating the Effects

Another possibility, perhaps the most popular at present, is simply to live with the effects of a warmer climate. After all, humans are extremely adaptable and already live with climate variations much greater than those that can be expected from the greenhouse effect. The changes will be phased in over a half century or so, and there will be considerable time to make necessary adjustments. Some observers see this as a more-or-less automatic process, while others believe that we should begin now to develop tactics for mitigating undesired effects.^[53]

What would be required to prepare for a drier climate? Especially in the midwestern and western United States, a small change in rainfall and runoff could lead to a large change in agricultural productivity unless adaptive measures are taken. While a fair amount is known about water conservation and re-use measures, a great deal more can be learned. Because planning the distribution of water for entire river basins requires a long lead time, the NRC has suggested that necessary research and modifications of water use begin in the near future. Ways to cope with a warmer climate might include: breeding genetic strains of crops that grow well in warmer and drier climates with higher CO₂ levels, stockpiling larger quantities of agricultural stocks to guard against famine, producing

114 

food in factories through methods such as fermentation, and farming in massive plastic greenhouses to conserve water.

Possible increase in sea level is another matter of concern. Hazardous waste dumps need to be located well out of the path of rising waters, and major new construction in coastal areas should have a relatively short coastal exposure and a sufficiently high proportion of valuable activities so that funds will be available for construction and maintenance of sea walls. Some research and development, on improved seawall construction for example, would also be necessary. Given the generally poor record for urban and regional planning, new federal laws may be required.

None of the above options have been examined in any detail, and no serious cost analyses have been attempted. The widespread assumption and it is only an assumption is that such measures would be less expensive than prevention of CO₂ buildup. As one prominent climatologist who advocates these and other adaptive strategies admits: "What can be done to prevent a tragedy of Dust Bowl proportions? The strategies are not entirely clear."^[54] So while the possibility of simply living with the greenhouse effect is an interesting option, its costs and benefits remain to be clarified.

Appraising the Current Strategy

The primary strategy employed to date against the greenhouse threat has been to employ a diverse number of physical scientists to study the risks. This research-and-monitoring approach is likely to remain the central strategy for the foreseeable future, with the goal of reducing the substantial uncertainties about the impact of the greenhouse threat. If agreement is reached on the need to take action, at least four strategies may be available:

Reducing the need for the risky activity by energy conservation, emphasizing nonfossil fuels, or finding safer alternatives to greenhouse gases;

115 

Restricting the risky activity by fuel taxes, production quotas for greenhouse gases, or selective bans on high-carbon fossil fuels;

Making the activities less risky, for instance, by offsetting the effects of fossil fuel use by reforestation;

Mitigating the effects of climate changes by altering patterns of agriculture, water use, and other activities likely to be affected.

Decisions about whether or when to employ these measures apparently are to be deferred pending further clarification of the greenhouse threat. What are the strengths and weaknesses of such attempts at clarification?

The Research-and-Monitoring System

The routes for focusing scientific attention in the greenhouse case were almost identical to those used in the ozone case. Early investigators conducted basic research on the atmosphere, glaciation, and other natural phenomena long before a greenhouse problem was suspected. As scientific knowledge gradually grew, scientists' concerns stimulated some government officials to increase funds for atmospheric research. This resulted in further improvements in the relevant sciences, together with authoritative reports on the subject of the possible greenhouse effect. These reports then became media events, which heightened public awareness and led to additional concern in Congress.

Four subcomponents of this process ordinarily taken for granted by those who participate in the process are important elements in an overall strategy that society has evolved for diagnosing potential catastrophe. First, a large and complex problem was broken down into many smaller, more comprehensible parts. For instance, glaciologists studied core samples of ice from Greenland to determine carbon dioxide levels thousands of years ago and used special dating and testing mechanisms developed by still other scientific specialists.^[55]

116 

Second, diverse subgroups of researchers competed to define the nature and extent of the problem. For example, computer simulations of climate by atmospheric chemists competed with interpretations of past geological experience. Oceanographers and atmospheric chemists attempted in very different ways to determine the percentage of released CO₂ that is absorbed into the oceans as compared with the atmosphere. Even within subfields, scientists at different universities and in different nations developed unique sets of data and approaches to interpreting such data.

Third, these disparate viewpoints are taken into account and to some degree reconciled by a standard process. One part of it is the normal scientific process: some ideas become dominant because they are more persuasive than the competing ones. In addition, when a major scientific organization such as the National Research Council produces a report, it typically gains the attention of media, policy makers, and the scientific community. The expressed view may be widely persuasive and lead to consensus, or it may be controversial and provide other scientists and policy analysts with a target against which to react in formulating their own versions of the situation. In the greenhouse case, such major committee reports contributed to a broader perspective than would the isolated work of individual scientists or policy analysts.

Fourth, government agencies typically commission further studies, which emphasize explicit policy concerns and help focus policy makers' attention on an issue. The EPA's controversial 1983 report, disavowed by the Reagan Administration, was the prime example to date in the greenhouse case. Its title clearly indicates the change in emphasis: "Can We Delay a Greenhouse Warming?"

Obstacles to Implementing a Strategy

The research-and-monitoring system described above is an effective method that will almost automatically diagnose threats to society from a variety of risky technologies. But diagnosis is only a first step; once the risk is perceived, further steps must be taken to guard against the worst possible out-

117 

come. These further steps have been taken for every other risky technology reviewed in this volume, but to date this has not occurred for the greenhouse problem. While it may be too early to expect action, we suspect that there are fundamental obstacles to effective action on this issue in the foreseeable future.

One of the largest obstacles is the ambiguity about whether the greenhouse effect would on balance be positive or negative. This ambiguity is not characteristic of any other risky technology reviewed in this volume (with the possible exception of the warm water released by nuclear power plants). Because some regions will benefit by the resulting changes in climate and precipitation patterns, the international agreement necessary to substantially reduce the greenhouse problem is difficult to achieve. Thus, the very nature of the greenhouse problem may prevent effective action to avert anticipated threats.

Another drawback in addressing this problem is the inability of scientists to narrow the remaining disagreements about the carbon dioxide controversy in a way that will enable policy makers to make decisions. While there is a working consensus among scientists regarding a buildup of carbon dioxide and other greenhouse chemicals, there is disagreement about how soon this will occur and how severe the effects will be. The range of uncertainty may render these calculations useless to policy makers. For instance, while the lower estimate of a 1.5 °C rise in mean global temperature might be widely acceptable (given the high costs of averting it), the highest credible estimated increase of 4.5 °C probably would be intolerable to the majority of policy makers. So unless scientists can pinpoint the risks more precisely, their findings may have little effect on policy. Yet in the period from 1979 through 1985 when investigations of the greenhouse effect probably tripled all previous research, virtually no progress was made in narrowing this range of uncertainty.^[56]

Given the potential risks, it is still reasonable to attempt to reduce remaining uncertainties through continued research and monitoring, but it is not clear how to proceed. The NRC provided a long list of research issues requiring further atten-

118 

tion, but it did not convey which, if any, questions were truly pivotal. Nor did it attempt to define the critical variables or uncertainties or to set priorities among research topics. What is necessary is not just more knowledge, but knowledge that will make a difference to government policy. Much more attention should be paid now to setting priorities for the research-and-monitoring system.

Another difficulty concerns the costs of action. While sustained research has been directed at illuminating the causes, timing, and magnitude of the greenhouse effect, much less effort has been expended in exploring possible actions to be taken in response to the anticipated problems. Whether to initiate action depends in large part on the costs of initiating action now versus the costs of postponing it. The policy sections of greenhouse studies typically assume in the absence of more compelling evidence that the costs of action are too large to justify policy changes. Most of the actions proposed so far would in fact be extremely costly, but too little attention has been given to searching for practical options.

For instance, there might be combinations of partial solutions that would keep costs bearable and also at least slow the increase of greenhouse gases. For example, an easing of the problem might be achieved by carefully researching how to adapt to temperature increases while at the same time limiting use of high-carbon fuels and instituting a partial reforestation program. So far, little effort has been made to explore such pragmatic solutions either in a national or international framework. The longer actions are delayed the more severe they may ultimately have to be.

Similarly, little effort has been devoted to studying the magnitude and timing of preventive actions. For example, if temperature increases are at the low end of the expected range, how long can we wait before taking action? On the other hand, if increases are at the high end, how much longer can we safely afford to wait?

Whatever information is obtained about costs and other aspects of the greenhouse effect, this will not automatically lead to a decision. Judgments about how much protection is desirable costs versus risks will still need to be made. In the

119 

cases of toxic chemicals and nuclear power, decision makers sometimes chose to err on the side of safety they moved ahead more slowly or designed more stringently than might have been strictly necessary. The equivalent approach in the greenhouse case would be to take initial steps to reduce or offset emissions of fossil fuels or greenhouse gases, even without conclusive evidence about the risks. But how soon would be soon enough to take such initial steps? Research and monitoring will not aid such a decision because there still remain conflicts over values.

One group of conservation/renewable energy advocates argues for action now:

To postpone action until climatic change is detected entails the risk of being unable to prevent further harmful changes that could prove irreversible for centuries.

. . . [Moreover] gradual changes are almost always more easily accomodated, in terms of both economic and social costs, than precipitous changes. . . . A lower growth rate of fossil fuel use over the next few decades, combined with a more efficient use of energy, would reduce the pressures for rapid societal and technological change later on and allow more time for development of alternative energy sources. Conversely, a more rapid increase in fossil fuel use during the next decade might necessitate an earlier and more drastic reduction.^[57]

In contrast, the NRC assessment committee was distinctly cool toward near-term action, as indicated in their 1983 report:

The potential disruptions associated with CO₂ -induced climatic change are sufficiently serious to make us lean away from fossil fuel energy options, if other things are equal. However, our current assessment . . . justifies primarily increased monitoring and vigilance and not immediate action to curtail fossil fuel use.^[58]

But this conclusion was based in part on scientific interpretations subsequently called into question. In particular, an important 1985 study reported the possibility that "most of the expected warming attributable to [already released] trace gases probably has not yet occurred." If there is a long delay before the earth's temperature fully adjusts to atmospheric

120 

changes, the scientists said, "this calls into question a policy of 'wait and see.'"^[59] Changes may be necessary now, for twenty-first-century changes in fossil fuel combustion and trace gas emissions might come too late to have much effect on climate until the following century.

The "How soon to act?" question is reminiscent of the "How safe?" issue that has plagued regulation of nuclear power. How conservative should society be in the face of gross factual uncertainties about the likelihood and magnitude of a technological danger? Waiting to act entails some risk; delay is not the most cautious action. Is waiting nevertheless justified? Are the risks of delay worth such benefits as minimizing economic repercussions or acquiring further information? There is no scientific answer to this question; it requires instead a political judgment. And once action is initiated (if it is), policy makers on the greenhouse problem will then be faced with the same question confronting nuclear power regulators: namely, how far should one go in protecting against possible risks?

We will have more to say about the "How safe? How soon?" question and about the costs of averting catastrophe in the concluding chapter.

Yet as clearly as the report recognizes the importance of the factual uncertainties, it fails to develop a strategy for dealing with them. It merely cites a long list of uncertainties that requires attention. As we discussed in chapter 6, the NRC listed over one hundred recommendations, ranging from economic and energy simulation models for predicting long-term CO₂ emissions, to modeling and data collection on cloudiness, to the effects of climate on agricultural pests.

Certainly answers to all of these questions would be interesting and perhaps useful; but, just as certainly, answers to some of them would be more important than answers to others. What are the truly critical uncertainties? What kinds of

167  

information would make the biggest differences in deciding whether or not to take action? As R&D proceeds and information is gained, are there key warning signals for which we should watch? What would be necessary to convince us that we should not wait any longer? Policy makers and policy analysts need a strategy for selectively and intelligently identifying, tracking, and reducing key uncertainties.

A similar problem arises in the case of nuclear power. In principle, nuclear regulators should systematically identify the central remaining safety uncertainties the issues that will continue to lead to new requirements for regulations. Regulators should then devise a deliberate R&D agenda to address such uncertainties. A prime example is uncertainty about the behavior of the reactor core once it begins to melt. Clearly, this lies at the heart of the entire nuclear debate, since the major threat to the public results from core melts. Yet, as we discussed earlier, virtually no research was performed on core melts in the 1960s and 1970s.

Information and research resulting from the experience of Three Mile Island now have called into question some of the basic assumptions about core melts. For example, if the TMI core had melted entirely, according to the Kemeny Commission it probably would have solidified on the containment floor.^[17] Even the nuclear industry had assumed that a melted core would have gone through the floor. Moreover, it appears that there were a variety of ways in which the core melt could have been stopped. Prior to the accident, the common assumption was that core melts could not be stopped once underway. Also overestimated, according to some recent studies, is the amount of radioactive material predicted to escape in a serious reactor accident: prior assumptions may have been ten to one thousand times too pessimistic.^[18]

If such revised ideas about reactor accidents were to be widely accepted, they would have a substantial effect on the perceived risks of reactor accidents. But all such analyses are subject to dispute. To the extent feasible, therefore, it clearly makes sense to invest in research and development that will narrow the range of credible dispute without waiting for the equivalent of a TMI accident. As with the greenhouse effect,

168  

what is needed is a systematic review of prevailing uncertainties and an R&D program devised to strategically address them. The uncertainties that make the biggest difference must be identified, those that can be significantly reduced by R&D must be selected, and an R&D program focused on these uncertainties must then be undertaken. In other words, a much better job can be done of using analysis in support of strategy.

Reduce Uncertainty
Improve Learning from Error

As noted previously, learning from error has been an important component of the strategies deployed against risky technologies. But learning from error could be better used as a focused strategy for reducing uncertainties about risk. As such, it would constitute a fourth strategic approach for improving the efficiency and effectiveness of the catastrophe-aversion system.

The nuclear power case again offers a good illustration of the need to prepare actively for learning from error. Suppose that a design flaw is discovered in a reactor built ten years ago for a California utility company. Ideally, the flaw would be reported to the Nuclear Regulatory Commission. The NRC would then devise a correction, identify all other reactors with similar design flaws, and order all of them to institute the correction. In actual operation, the process is far more complicated and the outcome far less assured.

To begin with, in any given year the NRC receives thousands of reports about minor reactor mishaps and flaws. The agency must have a method of sifting this mass of information and identifying the problems that are truly significant. This is by no means a straightforward task, as exemplified by the flaw that triggered the Three Mile Island accident. A similar problem had been identified at the Davis-Besse reactor several years earlier, but the information that was sent to the NRC apparently was obscured by the mass of other data received by the agency. Several studies of the TMI accident noted this unfortunate oversight, and concluded that the NRC and the nuclear industry lacked an adequate mechanism for monitoring feedback. In response, the nuclear industry established an

169  

institute for the express purpose of collecting, analyzing, and disseminating information about reactor incidents. This action represents a significant advance in nuclear decision makers' ability to learn from experience.

Even with a well-structured feedback mechanism, there are still other obstacles to learning from experience. One such obstacle arises from the contentious nature of current U.S. regulatory environments, which can actually create disincentives to learning. Given the adversarial nature of the nuclear regulatory environment, many in the nuclear industry believe that they will only hurt themselves if they propose safety improvements in reactor designs. They fear that opponents of nuclear power will use such safety proposals to argue that existing reactors are not safe enough, and that regulators will then force the industry to make the change on existing reactors, not just on new ones. This would add another round of costly retrofits.

Another obstacle to learning from experience can arise from the nature of the industry. For example, the nuclear industry is comprised of several vendors who over the years have sold several different generations of two different types of reactors to several dozen unrelated utility companies. Furthermore, even reactors of the same generation have been partially custom designed to better suit the particular site for which they were intended. This resulting nonuniformity of reactor design is a significant barrier to learning from experience, because lessons learned with one reactor are not readily applicable to others.

The design flaw uncovered at our hypothetical California utility's ten-year-old reactor probably can be generalized to the few reactors of the same generation (unless the flaw was associated with some site-specific variation of the basic design). It is less likely to apply to reactors built by the same vendor but of different generations, much less likely to apply to reactors of the same general type made by other vendors, and extremely unlikely to apply to other reactor types. Furthermore, lessons gained from experience in maintaining and operating reactors are also hard to generalize. Since reactors are owned by independent utilities, the experience of one util-

170  

ity in operating its reactor is not easily communicated to other utilities. In many respects, therefore, each utility must go through an independent learning cycle.

There also are significant barriers to learning about most toxic chemicals. The large number of such chemicals, the vast variety of uses and sites, and the esoteric nature of the feedback make the task of monitoring and learning from experience extraordinarily difficult. Yet the EPA's tight budget and the limited resources of major environmental groups means that routine monitoring will not get the attention that is given to other more pressing needs. What a good system for such monitoring would be is in itself a major research task, but just obtaining reliable information on production volumes, uses, and exposures would be a place to start.

The point, then, is that active preparation is required to promote learning from experience. The institutional arrangements in the regulatory system must be devised from the outset with a deliberate concern for facilitating learning from error. In the nuclear power case, the ideal might be a single reactor vendor, selling a single, standardized type of reactor to a single customer. The French nuclear system comes close to this pattern.^[19]

Conclusion

In summary, there are at least four promising avenues for applying risk-reduction strategies more effectively. The first strategy is to make an overall comparison of risks and to focus on those that clearly are disproportionate. The second is to transcend or circumvent risks and uncertainties by employing creative compromise, making technical corrections, and paying attention to easier opportunities for risk reduction. The third strategy is to identify key uncertainties and focus research on them. The fourth is to prepare from the outset to learn from error; partly this requires design of appropriate institutions, but partly it is an attitudinal matter of embracing error as an opportunity to learn. Finally, implicit throughout this study is a fifth avenue for improvement: by better under-

171  

standing the repertoire of strategies available for regulating risky technologies, those who want to reduce technological risks should be able to take aim at their task more consciously, more systematically, and therefore more efficiently.

Of these, the first strategy probably deserves most attention. Attacking egregious risks offers simultaneously an opportunity to improve safety and to improve cost effectiveness. As an example, consider the 1984 Bhopal, India, chemical plant disaster.^[20] The accident occurred when:

A poorly trained maintenance worker let a small amount of water into a chemical storage tank while cleaning a piece of equipment;

A supervisor delayed action for approximately one hour after a leak was reported because he did not think it significant and wanted to wait until after a tea break;

Apparently as an economy measure, the cooling unit for the storage tank had been turned off, which allowed a dangerous chemical reaction to occur much more quickly;

Although gauges indicated a dangerous pressure buildup, they were ignored because "the equipment frequently malfunctioned";

When the tank burst and the chemical was released, a water spray designed to help neutralize the chemical could not do so because the pumps were too small for the task;

The safety equipment that should have burned off the dangerous gas was out of service for repair and anyway was designed to accommodate only small leaks;

The spare tank into which the methyl isocyanate (MIC) was to be pumped in the event of an accident was full, contrary to Union Carbide requirements;

Workers ran away from the plant in panic instead of transporting nearby residents in the buses parked on the lot for evacuation purposes;

The tanks were larger than Union Carbide regulations specified, hence they held more of the dangerous chemical than anticipated;

172  

The tanks were 75 percent filled, even though Union Carbide regulations specified 50 percent as the desirable level, so that pressure in the tank built more quickly and the overall magnitude of the accident was greater.

The length of this list of errors is reminiscent of the Three Mile Island accident. The difference between the two incidents is that TMI had catastrophe-aversion systems that prevented serious health effects, while at least two thousand died in Bhopal and nearly two hundred thousand were injured. Even though the U.S. chemical industry is largely self-regulated, most domestic plants employ relatively sophisticated safety tactics that use many of the strategies of the catastrophe-aversion system. Still, questions remain about how effectively these strategies have been implemented.^[21] For example, a 1985 chemical plant accident in Institute, West Virginia, while minor in its effects, revealed a startling series of "failures in management, operations, and equipment."^[22]

The Bhopal and Institute incidents suggest that, relative to other risks, safety issues in chemical manufacturing deserve more governmental attention than they previously have received. In addition to whatever changes are warranted at U.S. chemical plants, special attention should be paid to the process of managing risk at many overseas plants owned by U.S. firms. If the practices at the Bhopal plant were typical, safety strategies abroad are haphazard. While the Bhopal incident has led to a fundamental review of safety procedures in chemical plants worldwide, it should hardly have required a catastrophe to reveal such a vast category of hazard. This oversight demonstrates that some entire categories of risk may not yet be taken into account by the catastrophe-aversion system.

The catastrophe-aversion system likewise was not applied, until recently, to hazardous waste in the United States. State and federal laws made no special provision for toxic waste prior to the 1970s; there were no requirements for initial precautions, or for conservatism in the amounts of waste that were generated. Systematic testing for underground contamination was not required, and waste sites were not monitored for potential problems. It is a tribute to the resilience of the

173  

ecosystem that after-the-fact cleanup now in progress has a good chance of keeping damage from past dumping below catastrophic levels. The next step is to find ways of limiting the generation of new wastes.

What does all this add up to? In our view, society's standard operating procedure should be as follows:

First, apply each of the catastrophe-aversion strategies in as many areas of risk as possible;

After this has been accomplished, proceed with more detailed inquiry, debate, and action on particular risks.

To pursue detailed debates on a risk for which a catastrophe-aversion system already is operative, continuing to protect against smaller and smaller components of that risk, is likely to be a misallocation of resources until the full range of potential catastrophes from civilian technologies has been guarded against. The "How safe?" questions that have become so much the focus of concern are matters of fine tuning; they may be important in the long run, but they are relatively minor compared to the major risks that still remain unaddressed.

Concluding Note

At the outset of this volume we quoted a highly respected social critic, Lewis Mumford, who claimed in 1970 that "The professional bodies that should have been monitoring our technology . . . have been criminally negligent in anticipating or even reporting what has actually been taking place." Mumford also said that technological society is "a purely mechanical system whose processes can neither be retarded nor redirected nor halted, that has no internal mechanism for warning of defects or correcting them."^[23] French sociologist Jacques Ellul likewise asserted that the technological

system does not have one of the characteristics generally regarded as essential for a system: feedback. . . . [Therefore] the technological system does not tend to modify itself when it develops nuisances or obstructions. . . . [H]ence it causes the increase of irrationalities.^[24]

174  

Reflecting on different experiences several decades earlier, Albert Schweitzer thought he perceived that "Man has lost the capacity to foresee and forestall. He will end by destroying the earth."^[25]

Although one of us began this investigation extremely pessimistic and the other was hardly an optimist, we conclude that Mumford, Ellul, Schweitzer, and many others have underestimated the resilience both of society and of the ecosystem. We found a sensible set of tactics for protecting against the potentially catastrophic consequences of errors. We found a complex and increasingly sophisticated process for monitoring and reporting potential errors. And we found that a fair amount of remedial action was taken on the basis of such monitoring (though not always the right kind of action or enough action, in our judgment).

Certainly not everyone would consider averting catastrophe to be a very great accomplishment. Most citizens no doubt believe that an affluent technological society ought to aim for a much greater degree of safety than just averting catastrophes. Many industry executives and engineers as well as taxpayers and consumers also no doubt believe that sufficient safety could be achieved at a lower cost. We agree with both. But wanting risk regulation to be more efficient or more effective is very different from being caught up in an irrational system that is leading to catastrophic destruction. We are glad and somewhat surprised to be able to come down on the optimistic side of that distinction.

Finally, what are the implications of the analysis in this volume for environmentally conscious business executives, scientists, journalists, activists, and public officials? Is it a signal for such individuals to relax their efforts? We do not intend that interpretation. The actions taken by concerned groups and individuals are an important component of the catastrophe-aversion system described in these pages. To relax the vigilance of those who monitor errors and seek their correction would be to change the system we have described. Quick reaction, sometimes even overreaction, is a key ingredient in that part of regulating risky technologies that relies on trial and error. So to interpret these results as justifying a reduction of efforts would be a gross misreading of our message.

175  

Instead, we must redirect some of our concern and attention. Environmental groups should examine whether they could contribute more to overall safety by focusing greater attention on egregious risks that have not been brought under the umbrella of the catastrophe-aversion system instead of focusing primarily on risks that already are partially protected against. The Union of Concerned Scientists, for example, devotes extended attention to analyses of nuclear plant safety but has contributed almost nothing on the dangers of coal combustion, international standards for chemical plants, or toxic waste generation egregious risks that have not been taken into account by catastrophe-aversion strategies. Regardless of whether contemporary nuclear reactors are safe enough, there is no question that they have been intensively subjected to the restraints of the catastrophe-aversion system. We doubt that much more safety will be produced by further debate of the sort that paralyzed nuclear policy making during the 1970s and 1980s. In general, we believe it is time for a more strategic allocation of the (always limited) resources available for risk reduction.

The main message of this volume, however, has been that the United States has done much better at averting health and safety catastrophes than most people realize, considering the vast scope and magnitude of the threats posed by the inventiveness of science and industry in the twentieth century. Careful examination of the strategies evolved to cope with threats from toxic chemicals, nuclear power, recombinant DNA, ozone depletion, and the greenhouse effect suggests that we have a reasonably reliable system for discovering and analyzing potential catastrophes. And, to date, enough preventive actions have been taken to avoid the worst consequences. How much further improvement will be achieved depends largely on whether those groups and individuals concerned with health and safety can manage to win the political battles necessary to extend and refine the strategies now being used. Because we have a long way to go in the overall process of learning to manage technology wisely, recognizing and appreciating the strengths of our catastrophe-aversion system may give us the inspiration to envision the next steps.