LL1-17: Ch17 Conclusions#

Report: Late lessons from early warnings: the precautionary principle 1896–2000 (EEA Environmental Issue Report No 22, 2001) Report pages: 192–194. PDF pages: 192–194 (in this report the printed and PDF page numbers are the same) Structure: an untitled opening (pp. 192–193), including Table 17.1 (p. 192); then §17.1 “Late lessons from early warnings” (pp. 193–194).

What was read. I read the whole text extract (working/text/chunks/LL1-17.txt, PDF 192–194) and confirmed the final page marker (PDF 194). I rendered all three pages from the PDF and checked them against the extract. The extraction is clean. Table 17.1 came through correctly. The only formatting lost is the italic Late lessons (the report title) on p. 193 and the bold action labels (“Prevention”, “Precautionary prevention”, “Precaution”) in the table’s third column. Ch17 is a three-page synthesis that leans on earlier chapters, so I also read these parts of the same PDF for context and cross-checking: Preface (pp. 3–5), Contents (pp. 7–10), Introduction (pp. 11–16), Ch16 “Twelve late lessons” (pp. 168–189), the case tables of warnings and actions for asbestos (Table 5.1, p. 61), halocarbons (Table 7.1, p. 83) and antimicrobials (Table 9.2, p. 99), and the author biographies (pp. 195–199). Material from outside Ch17 is always labelled as such.

Note on quotation. Verbatim quotation is kept deliberately short: key phrases and terms of art only. The exact wording of the twelve lessons and of Table 17.1 is in the local extract (LL1-17.txt, lines 68–98 for the table and 197–240 for the lessons) and on pp. 192–194 of the PDF. Below, the lessons and the table are closely paraphrased, with the report’s own terms kept.

Authors and standpoint#

Authorship. The chapter carries no byline. The Contents (p. 10) list “17. Conclusions” without an author, while Ch16 “Twelve late lessons” is credited to the “Editorial team” (pp. 10, 168). The Preface (p. 3) says the case authors’ lessons were distilled into twelve “late lessons” by the editorial team under the guidance of the EEA Scientific Committee. The most reasonable reading is that Ch17 speaks for the editorial team collectively and is the report’s institutional conclusion. Which individuals drafted it is not recorded. The Preface is signed by the EEA Executive Director, Domingo Jiménez Beltrán (p. 5), not by the editors. The EEA Scientific Committee members thanked on p. 6 are Philippe Bourdeau, Eileen Buttle, Robert Kroes, Bo Jansson and Poul Harremoës.

Editorial team (title page p. 1; biographies pp. 195–199): - Poul Harremoës (Chair). Professor of environmental science and engineering at the Technical University of Denmark; member of the EEA Scientific Committee. Background in water engineering (urban water, water pollution; former president of the International Water Association, p. 196). He co-authored the MTBE chapter (Ch11) with Martin Krayer von Krauss, a PhD student in his DTU department (Contents p. 8; biographies pp. 196, 199). This matters because Ch16 and these notes lean heavily on MTBE as an example. - David Gee (EEA editor; co-project manager; the Preface, p. 5, says he initiated the report). A graduate in economics and politics who had worked on occupational and environmental health since 1974 for trade unions and NGOs, and a former Director of Friends of the Earth (England, Wales and Northern Ireland). At the EEA since 1995 on emerging issues. He co-authored the asbestos chapter. - Malcolm MacGarvin (Executive editor). PhD in ecology; consultant to industry, NGOs, the European Commission and the EEA. Author of the fisheries chapter. - Andy Stirling (Editor). Senior Lecturer at SPRU, University of Sussex, working on technology assessment, appraisal and risk. Ch16 (p. 168) says the ESTO project on technological risk and the management of uncertainty (Stirling, 1999) “provided the initial framing of this analysis”. Ch16’s reference list (pp. 190–191) records that the same Stirling (1999) ESTO report was cited in the European Commission’s Communication on the Precautionary Principle (COM(2000)1). That shows a shared intellectual lineage between the report’s lessons and the Commission text whose trigger language lesson 12 echoes. It does not show that the ESTO work supplied the Communication’s “reasonable grounds for concern” wording; the report makes no such claim. - Jane Keys (Editor). Freelance researcher for industry, NGOs and the EEA. Co-author of the PCBs chapter. - Brian Wynne (Editor). Professor of Science Studies at Lancaster; member of the EEA management board and Scientific Committee 1995–2000. Known for work on public understanding of science, risk and ignorance (Ch16 cites Wynne 1992, 2001). - Sofia Guedes Vaz (EEA editor; co-project manager). Environmental engineer at the EEA since 1997.

Evident stance. The chapter argues for precaution but presents itself as seeking balance. The word “balance” recurs on pp. 192 and 194, and the chapter concedes that over-precaution has costs (p. 194). Its concepts come from a recognisable intellectual school. Science and technology studies, and Stirling’s appraisal framework in particular, supply the risk/uncertainty/ignorance typology, the stress on public values and “framing”, and the claim that inclusion yields better science. That school shaped the conclusions alongside the case evidence. Ch16 says the ESTO framework “provided the initial framing” and that the historical material gave “an opportunity to test or elaborate many of the points arising from the ESTO studies” (p. 168). In the same passage it says “most of the key issues that emerged from the case studies” could be addressed by the twelve lessons, and later that the lessons are “well grounded in the empirical detail” of the cases (p. 182). How much weight each carried cannot be measured from the text.

Institutional standpoint. The EEA describes itself as an information agency whose duty is to provide objective information to EU policy-makers (Ch1, p. 11). Ch17 describes most of the lessons as improvements to information (p. 194), which fits that mandate (see the bias check).

Political context (from the Preface and Introduction, not Ch17 itself). The chapter was written in 2001: - after the European Commission’s Communication on the Precautionary Principle (COM(2000)1) and what Ch1 calls the “Council of Ministers Nice Decision”, both in 2000 (Ch1, p. 13; Ch16, p. 186 calls it the Nice “conclusions”). The months (February and December 2000) come from outside the report; - during EU–US disputes over hormones in beef, GMOs and climate (Ch1, p. 12); - in a crisis of public trust after BSE, the Belgian dioxin affair and France’s HIV-contaminated blood scandal (Ch1, p. 16); - alongside the EU White Paper on European Governance of July 2001 (Ch1, p. 16).

One stated aim of the report is better transatlantic understanding of precaution (Preface, p. 3; Ch1, p. 12).

Panels and commentaries. None. The 2001 report has no panels (they arrive in the 2013 volume), and Ch17 records no dissent. The report’s general disclaimer (p. 2) says its contents do not necessarily reflect the official opinions of the European Commission or other EU institutions.

Section-by-section notes#

1. Two basic observations (p. 192)#

The chapter opens by saying that learning from history begins with two observations.

Observation 1: regulation is a balancing act under uncertainty and ignorance. Regulatory appraisal and control of technologies and economic development means weighing the costs of being too restrictive on innovation against the hazards and costs of being too permissive. The cases show many examples of regulatory inaction producing costly consequences that were not foreseen. The chapter adds that some of these consequences could not have been foreseen at all. - Analytic note. The concession matters. The report does not claim that every harm was foreseeable. That leaves room for the “ignorance” category (Table 17.1) and for strategies aimed at surprise itself rather than at specific hazards.

Observation 2: three kinds of failure. The cases also show: 1. early warnings, and even “loud and late” warnings, that were clearly ignored; 2. hazard appraisal whose scope was too narrow; 3. regulatory actions taken without enough thought about alternatives, or about the conditions needed to make them work in the real world. - Analytic note. The third point is often overlooked. The report criticises badly designed regulatory action as well as inaction. Ch16 supplies the examples: MTBE was adopted partly as the solution to lead in petrol, with its persistence and taste and odour problems in groundwater overlooked (Ch16, p. 176). Taller smokestacks solved local sulphur dioxide episodes but pushed the problem into long-range acidification (Ch16, pp. 174, 176).

A richer information base. If society takes more account of a richer body of information from more diverse sources, scientifically, politically and economically, it “may do substantially better” at balancing innovations against their hazards (p. 192). The twelve lessons came out of “discussion of the case studies” (p. 192). - Ch16 (pp. 168–169) is franker about method. The lessons were organised around a pre-existing framework (ESTO), and the case material gave “an opportunity to test or elaborate” many of its points.

Definition of the precautionary principle. The chapter calls it an overarching framework of thinking that governs the use of foresight where uncertainty and ignorance prevail and where both regulatory action and inaction may carry large costs (p. 192). - Analytic note. This is a deliberately broad definition: a way of thinking rather than a trigger for decisions. It contrasts with the EC Communication’s more operational framing and with narrow “trigger” formulations such as Rio Principle 15 (quoted in Table 1.2, p. 14). It does not contrast with Ch1, which already says a precautionary approach “requires much more than” setting the level of proof (the “trigger”) and lists the German Vorsorgeprinzip’s components (p. 13): research and monitoring, reducing burdens, clean production and innovation, proportionality, cooperation, and action before full proof. Ch17’s definition generalises that broad reading, and Ch16 closes by saying the principle, “once accepted, leads far beyond the simple definition” (p. 189). The breadth has two effects. The principle can absorb innovation, participation and science reform. It also becomes harder to pin down, and harder for critics to engage.

The terminology problem. Applying the principle to complex, uncertain and controversial issues is made harder by a lack of agreement on key terms (p. 192). Table 17.1 is offered as a contribution to clarifying six basic concepts. What is loosely called “uncertainty” mixes three analytically distinct states of knowledge (risk, uncertainty, ignorance). Three public-action concepts (prevention, precautionary prevention, precaution) are mapped onto them.

2. Table 17.1: “Uncertainty and precaution — towards a clarification of terms” (p. 192; source: EEA)#

Close paraphrase (verbatim text at extract lines 68–98):

Analysis of the table

1. Knowledge states are dated. The second column is headed “State and dates of knowledge”. The same hazard moves between categories over time: asbestos mesothelioma is “ignorance” before 1959 and “risk” from 1965. A state of knowledge belongs to a historical moment and to the institutions of that moment, not to the hazard. The table implies this without saying it. Note also that the table leaves 1959–1965 unassigned for mesothelioma. By its own logic those years would presumably be “uncertainty” (impact known, probabilities not), but the table does not say so.

2. The dates, checked against the report’s own case tables. - Asbestos. Table 5.1 (p. 61) has mesothelioma identified in South Africa in 1959–60 and in the UK and US in 1962–64. Ch5 (p. 55) says most experts accepted causation by 1964. Ch5 (p. 55) also records that mesothelioma cases “had been observed in association with asbestos exposure in the 1940s and 1950s”, and that a local South African doctor (Sleggs) noticed a cluster in 1955, before Wagner and colleagues published in 1960. So “ignorance before 1959” is itself a simplification: scattered observations existed earlier. The asbestos chapter’s Dutch estimate compares a hypothetical 1965 ban, “after the mesothelioma evidence had been widely accepted”, with the actual 1993 ban (p. 58). So 1965 marks consensus on mesothelioma. - Table 5.1 also records the Merewether report of 1930, which found 66% of long-term workers at a Rochdale factory had asbestosis; the 1931 UK regulations; and Doll’s 1955 finding of high lung cancer risk. Dating “respiratory disease, lung and mesothelioma cancer” as a known risk from 1965 therefore understates how long asbestosis and lung cancer had been known. On the report’s own evidence, the “risk” state for asbestosis goes back at least to 1930. The compression understates the report’s own case about delayed prevention. - Antimicrobials. The 1969 date is the UK Swann Committee report (Table 9.2, p. 99; Swann, 1969, cited p. 94). Table 9.2 also records Fleming’s 1945 warning about misuse of penicillin, vertical transmission of resistance recognised in the 1950s and horizontal transmission in the 1960s. The report is partly inconsistent on the Swann dates. The antimicrobials chapter says the committee was established in 1968 and reported in 1969 (Ch9, p. 94). On that account Ch16’s evidence “available in 1968” (p. 173) and “the UK Swann Committee in 1969 concluded” (p. 181) fit together. Two other Ch16 phrasings do not fit: “the 1967 Swann Committee” (p. 172) and an “early 1968 recommendation” (p. 174), which would predate the committee’s own setting-up. - CFCs. The 1974 date is Molina and Rowland (Table 7.1, p. 83). Table 7.1 also lists laboratory experiments on the decomposition of ozone photosensitised by chlorine in 1907 (Weigert) and 1934 (Norrish and Neville), and Lovelock’s 1973 global survey showing CFCs distributed worldwide. The Preface (p. 3) concedes that “some may argue that important clues were missed earlier”. The report’s own considered view, however, is that pre-1974 CFCs are “a very strong candidate for ignorance extending over many decades”: “the very possibility of an ‘ozone hole’ was unappreciated” (Ch16, p. 169). The CFC chapter’s author (Farman) also frames it as a move from ignorance to “understanding, however rudimentary” (Ch7, p. 83). The 1907 and 1934 experiments were general laboratory chemistry. They did not show that CFCs would reach and act in the stratosphere. My reading: the “ignorance” label is defensible on the report’s evidence. A weaker version of the objection survives: some component knowledge existed before 1974 but had not been put together, which is closer to what Ch16 (p. 171) calls “institutional ignorance” (knowledge in society but not at the point of decision) than to pure “societal ignorance”. Table 17.1 does not carry that distinction.

3. The “precaution” row is not tied to any specific harm, and this is the conceptual core. Its strategies do not require knowing what the harm will be: screen on intrinsic properties such as persistence and bioaccumulation, cast the information net wide, monitor over the long term, and keep technological and social options diverse and adaptable. This is how the report answers the objection that one cannot act on what one does not know. Ch16 elaborates, with its own hedges (“although not quite as simple as it seems”, p. 170; the case studies suggest it is possible “to do rather better than in the past”, p. 170): - pp. 170–171: account can be taken of the potential irreversibility of actions “even if the consequences might not be known”. The novelty, persistence and ready dispersal of artificial chemicals (halocarbons, PCBs, MTBE) could each have served as a warning, and screening on persistence and bioaccumulation should make the “size and seriousness” of future “surprises” smaller. The scale of a potential hazard, especially a global one “where there is only one ‘experimental’ model”, is also relevant (p. 171). - p. 171: Ch16 assigns different remedies to its two kinds of ignorance. “Institutional ignorance” calls for better communication and social learning. “Societal ignorance” calls for research and for “greater diversity, adaptability and flexibility in decision-making and technological choices”. - p. 187: diversity acts as “insurance”, because a surprise is smaller when several competing technologies meet a need rather than one global near-monopoly (asbestos, halocarbons, PCBs). - Property-based triggers were not new in 2001. The Third North Sea Conference (1990) had already committed to act on substances that are “persistent, toxic, and liable to bioaccumulate” even without proof of a causal link (Table 1.2, p. 14). Ch16 (p. 183) says Swedish chemicals policy uses persistence and bioaccumulation as “proxies” for unknown impacts.

4. The typology is a simplification of the editors’ own. Ch16 Box 16.1 (p. 170) defines risk as in probability theory, uncertainty as the absence of an adequate basis for assigning probabilities, and ignorance as the condition in which some possibilities are themselves unknown. Ch16 (p. 184) then notes further dimensions: complexity, indeterminacy, ambiguity and the nature of disagreement (citing Wynne 2001 and Stirling 1999). Table 17.1 leaves out ambiguity, meaning disagreement about framing and values. Yet p. 193 goes on to describe current controversies as being about values, so the three-way typology does not fully cover the conflicts the chapter itself highlights. In fairness, the editors do not ignore values: Box 16.1 (p. 170) says judgements are “laden with subjective assumptions and values” even under risk, and Ch16 (p. 186) cites Popper that policy cannot be derived “from facts alone”. The gap is that Table 17.1 does not formalise these points, not that the report overlooks them.

5. “Precautionary prevention” does diplomatic work. Ch1 uses the term for John Snow’s removal of the Broad Street pump handle in 1854 (p. 14) and for United States practice (Table 1.1, p. 12): the Delaney Clause, the 1970s ban on scrapie-infected meat, the 1977 ban on CFCs in aerosols, and the 1972–79 ban on DES as a growth promoter. The middle category lets the report show that the US practises precaution without using the word, which fits its transatlantic aims. This is stated rather than implied: Ch1 (p. 12) says the US “has helped to promote what could be called ‘precautionary prevention’, without necessarily calling it ‘the precautionary principle’”. Ch1 (p. 13) also roots the term in medicine and public health (“better safe than sorry”).

Procedural requirement (pp. 192–193). Procedures for handling risk, uncertainty and ignorance should be fair, transparent and accountable. These are key elements of the “good governance” needed to regain public confidence in policy on technologies, their benefits and their hazards. This echoes Ch16 Box 16.1 (p. 170), where “robust, transparent and accountable” approaches to risk, uncertainty and ignorance are called “one crucial means of regaining public confidence”.

3. Integrating health and environment (p. 193)#

Most of the cases involved costly impacts on both public health and the environment. Over the past century these became specialised and “somewhat polarised” fields of science and policy. People experience their health and their environment as one reality, so science, regulatory appraisal and policy should be integrated in the same way. The report hopes to contribute to that. - The evidence is in Ch16 (lesson 4, p. 174): - standard-setting for asbestos and radiation was dominated by clinicians’ focus on acute effects; - MTBE was appraised through knowledge of engines, combustion and air pollution, while its water pollution was disregarded; - the sulphur dioxide regime was built around human health and struggled to absorb ecological effects; - for livestock antimicrobials and BSE, the human impacts were marginalised by a veterinary focus. - Ch16 (p. 171) adds that harm seen in wildlife can serve as a “sentinel” for human hazards. - On “most of the cases”, my assessment is that the claim is defensible but loose. Benzene, DES, radiation and (for the most part) asbestos are chiefly human-health cases, and fisheries and TBT chiefly environmental, so “most” rests on a modest majority of the 14 (PCBs, halocarbons, sulphur dioxide, MTBE, the Great Lakes, antimicrobials, hormones and arguably BSE). Ch17 gives no count.

4. Widening appraisal; involving stakeholders early (p. 193)#

• Regulatory appraisal should be widened to give adequate weight to relevant social issues alongside the physical, chemical, biological and medical aspects of technologies.
• Involving a wide range of stakeholders, and taking account of their values and interests, at the earliest stage of appraising and choosing among technological and social options for meeting human needs, brings two benefits: 1. it adds to the information available to policy-makers; 2. it may improve public trust in society’s capacity to control hazards, “without necessarily” stifling innovation or compromising science.
• Note the hedges (“may”, “without necessarily”).
• Note also the frame. The object of appraisal is options for meeting needs, not a particular product. This ties into lesson 7 on alternatives.
• Ch16 (p. 186) grounds early involvement in the US National Research Council’s Understanding Risk (1996), the UK Royal Commission on Environmental Pollution (1998) and the Nice conclusions (2000). It argues that appraisal, management and communication are not sequential stages. (These sources are introduced in §16.3, pp. 182–186, which by the editors’ own account steps outside the case material; see §8 below.)
• Ch16 attaches caveats that Ch17 leaves out. Early involvement should not mean “paralysing indiscriminate full public participation in every single decision”. A consensus conference without safeguards “can potentially degenerate” into consultation driven by the sponsor’s agenda. The tools for participation are still developing and “the challenges are far from trivial” (all p. 188). Lay knowledge also needs the same critical scrutiny as specialist expertise (p. 178).

5. The changing character of controversy (p. 193)#

• Recent controversies over “emerging technologies such as genetically modified organisms and oil-rig disposal” are said to have much to do with public values and scientific uncertainties. This is contrasted with an earlier period, in which values carried little weight and unequivocal scientific proof was demanded before action to reduce hazards.
• Openly recognising this changed context is described as a first step towards better governance of such controversies.
• Notes:
• “Oil-rig disposal” almost certainly refers to Brent Spar (1995); Ch17 does not name it. Ch16 (p. 188) names the “abandoned attempt to dump” it, together with public rejection of irradiated food and the response to GMOs, as costs of failure under traditional approaches. Calling rig disposal an “emerging technology” is loose.
• Neither GMOs nor Brent Spar is a case study in the report, so this paragraph is interpretation from outside the case evidence.
• The historical claim, that the demand for unequivocal proof before action was once dominant, is well supported by the report’s own cases (benzene, asbestos, PCBs, acid rain). The sentence’s syntax (“in contrast to the previously low emphasis on values and to the demand for unequivocal scientific proof”) could also be read as saying the demand for proof persists. The reading above, that both belong to the earlier period, is the more natural one.
• Ch16 records the transatlantic contestation that Ch17 does not: EU risk assessment of GMOs is criticised by some specialists as too reductionist and narrow, while “US counterparts” regard it as “irrationally exaggerated” by their standard of “sound science” (p. 183).

6. Better science through inclusion, and complexity (p. 193)#

• Including different socio-political perspectives in regulatory appraisal is said to make for better science as well as better policy.
• The example given: reductionist science and linear causality are useful but limited. They cope poorly with complex and sometimes chaotic systems that feature feedback loops, synergisms, thresholds, equilibrium and instability, and multi-factorial, interdependent causal chains. Such a reality demands better science, with “more humility and less hubris” and attention to what we don’t know as well as to what we do.
• Where the phrase comes from. “More humility and less hubris” appears in the antimicrobials chapter by Edqvist and Pedersen (Ch9, p. 98). Ch16 quotes it (p. 170) in their criticism that the Swann recommendations were diluted on the basis of what was precisely known rather than what was not. Ch1 (p. 15), which carries no byline, also uses it, and there too in connection with “the emerging science of complexity and chaos”. Which use came first cannot be told from the report. What is clear is that the phrase runs through the report’s framing chapters as well as one case chapter.
• A logical gap in the compression. The example (complexity) does not show how including socio-political perspectives improves science. The linking arguments are in Ch16:
• lay knowledge is often better grounded in real operating conditions (pp. 177–178);
• the intuitions in public values “may sometimes prove quite robust in relation to the framing of the regulatory science itself”, as with public revulsion at feeding offal to ruminants (BSE) and public misgivings about antimicrobials in animal husbandry (p. 178);
• members of the public distinguish uncertainty from ignorance, and institutions deny ignorance (p. 185).

Without those links the Ch17 sentence reads as assertion. - A caveat Ch17 omits. Ch16 (pp. 172–173) concedes that research can compound uncertainty and reveal new ignorance. A Canadian multi-species fisheries model became less predictable as more biological data were added, and intensive Great Lakes research amplified uncertainty about the causes of bird population crashes. So more science is not automatically precautionary. Ch16 also concedes that where research shows a concern to be unfounded, “it is in no way precautionary to persist in restricting the original agent” (p. 173). It then offers a reply (p. 181). Broadening prospective appraisal (from agent to possible effects) is precautionary. Broadening retrospective appraisal (from observed effects back to possible agents) may raise uncertainty about any one agent. Even so, the principle may “entirely legitimately” be invoked to keep action on that agent until the uncertainty is resolved.

7. Levels of proof (p. 193)#

• If one accepts both scientific uncertainty (including ignorance) and the urgency of reducing hazards when stakes are high, then agreement is needed on how much evidence of harm justifies action.
• These “levels of proof” range from the European Commission’s “reasonable grounds for concern” to the criminal law’s “beyond reasonable doubt”.
• Choosing a level for a given situation can radically shift the size, nature and distribution of the costs of being wrong. The chapter calls this a key political decision with profound ethical implications.
• The right level depends on:
• the size and nature of the potential harm;
• the claimed benefits;
• the available alternatives;
• the potential costs of being wrong in both directions, whether by acting or not acting, under uncertainty, ignorance and high stakes.
• An analogy is offered: military intelligence has long taken similarly precautionary approaches where the costs of error can be catastrophic. No source is given.

Verification and context: - The EC Communication (COM(2000)1). Its summary does use “reasonable grounds for concern”, in the sentence describing when recourse to the principle is warranted (checked on EUR-Lex). The Communication also sets out principles for applying precaution: proportionality, non-discrimination, consistency, examination of the benefits and costs of action and inaction, and review. It discusses burden of proof separately (§6.4). - Ch16’s Table 16.1 (p. 184) gives the fuller ladder: - “beyond all reasonable doubt”: criminal law, and the evidence of safety required from manufacturers under Sweden’s 1973 chemicals law; - “balance of evidence”: the IPCC in 1995 and 2001; - “reasonable grounds for concern”: the EC Communication; - “scientific suspicion of risk”: the evidence regulators need before acting under the same Swedish law. - Ch17’s range is narrower than Ch16’s. Ch17 gives “reasonable grounds for concern” as the bottom of the range and “beyond reasonable doubt” as the top (p. 193). Box 16.1 (p. 170) and Table 16.1 (p. 184) list a still lower bar (“scientifically based suspicion” / “scientific suspicion of risk”), and Table 16.1 words the top one as “beyond all reasonable doubt”. Dropping the lowest rung makes the EC Communication’s standard look like the most precautionary option on offer, which the editors’ own table shows it is not. - The Swedish example sets a high bar for manufacturers to prove safety and a low bar for regulators to act. That is really a way of allocating the burden of proof, a point Ch17 does not spell out. - Ch16 (pp. 183–184) puts the statistical case. Scientific convention guards against Type I errors (false positives) and neglects Type II errors (false negatives), so “not being wrong” is prized over “being safe”. Low statistical power and misclassified exposure make so-called “negative” studies falsely reassuring. Ch17 compresses all of this into the levels-of-proof paragraph. - Box 16.1 (p. 170) already makes the distributive point: each level of proof has “different cost and benefit implications for different groups”. - A case-chapter concession. The asbestos chapter (Gee and Greenberg, Ch5, p. 60) accepts that shifting science’s bias from avoiding false positives towards “a better balance” would “increase the chances of generating the costs of restricting a substance or activity that might later turn out to be safe”. It argues that society would still gain overall. Ch17’s “both directions” wording is consistent with this, but Ch17 does not state the trade-off so plainly. - Assessment. This is the chapter’s most rigorous passage. The cost of error is framed symmetrically (“in both directions”), which answers the caricature of precaution as acting on any suspicion. The chapter says the choice is political, and elsewhere that procedures must be fair, transparent and accountable (pp. 192–193). It does not say which body should set the level in a given case, by what procedure, or how the four factors are to be weighed against one another.

8. §17.1 Late lessons from early warnings: the twelve lessons (pp. 193–194)#

The chapter says the cases both support and illustrate the need for twelve lessons drawn from a century of history. Close paraphrase (verbatim at extract lines 197–240):

1. Acknowledge and respond to ignorance, not only uncertainty and risk, in technology appraisal and public policy.
2. Provide adequate long-term environmental and health monitoring, and research into early warnings.
3. Identify and work to reduce “blind spots” and gaps in scientific knowledge.
4. Identify and reduce interdisciplinary obstacles to learning.
5. Ensure regulatory appraisal adequately accounts for real-world conditions.
6. Systematically scrutinise the claimed justifications and benefits alongside the potential risks.
7. Evaluate a range of alternative options for meeting needs alongside the option under appraisal. Promote more robust, diverse and adaptable technologies so as to minimise the costs of surprises and maximise the benefits of innovation.
8. Use “lay” and local knowledge as well as relevant specialist expertise.
9. Take full account of the assumptions and values of different social groups.
10. Maintain regulatory independence from interested parties while keeping an inclusive approach to gathering information and opinion.
    • The printed wording is “independence of interested parties”, which is ambiguous drafting. The Ch16 heading (p. 178) makes the meaning clear: independence from economic and political special interests.
11. Identify and reduce institutional obstacles to learning and action.
12. Avoid “paralysis by analysis” by acting to reduce potential harm when there are reasonable grounds for concern.

Notes on the list: - Relationship to Ch16. The list is word for word the same as in Ch16 (pp. 168–169). Ch16 adds caveats that Ch17 drops: - the distinctions are “illustrative, rather than definitive”, many are interlinked, and some “might be combined together or further distinguished” (p. 169); - many lessons concern information “set within the context of a more participative and democratic process”, and the appraisal process “needs to be related to the likely scale of the potential consequences” (p. 169). Ch17’s parallel sentence on information (p. 194) keeps the first half and drops both qualifiers; - one “should avoid … over-reliance on any single set of prescriptions” for what is precautionary (p. 183); - the lessons were elaborated under a rule of not introducing material from outside the case studies, but “a final section” (§16.3, pp. 182–189) deliberately sets them in wider context (p. 169). Much of the support these notes draw from Ch16 on innovation, participation, diversity and statistics comes from that section, so it is the editors’ wider argument rather than case evidence. Ch16 (p. 182) also sets out the criteria the lessons were built to meet: grounded in the empirical detail of the cases, general enough to apply to virtually any risk problem, balanced and fairly comprehensive together, and concrete enough to guide practice. - Grouping (mine): - producing and handling knowledge (1–5); - scope of appraisal: benefits and alternatives (6–7); - participation and values (8–9); - interests and institutions (10–11); - the trigger for action (12). - Lesson 12 uses the same trigger language as the EC Communication, which aligns the report with EU policy of the time. (The shared ESTO lineage is noted under Authors above.) - Tensions inside the list. - Lessons 1–9 largely say “know more”. Lesson 12 says “don’t wait”. Ch16 (pp. 181–182) acknowledges this tension and says that whether a demand for more information counts as prudence or as paralysis depends on how the pros and cons fall on those making the judgement. - Lesson 10 (independence) sits uneasily with lessons 8–9 (inclusion). The wording of lesson 10 tries to hold both.

9. Closing claims (p. 194)#

• Information. Most lessons concern improving the quality, availability, use and processing of information in environmental and health policy.
• Limits. None of the lessons would remove the dilemmas of deciding under uncertainty with high stakes. They cannot eliminate uncertainty or avoid the consequences of ignorance.
• Expected benefits. They would “at least increase the chances” of three things: anticipating costly impacts, striking a better balance between the pros and cons of innovations, and minimising the costs of unpleasant surprises.
• Benefits beyond harm reduction. Using the precautionary principle can also stimulate more innovation, through technological diversity and flexibility, and better science.
• Balance. The lessons may help balance “proportionate and precautionary” policies, since over-precaution can also be expensive through lost opportunities for innovation and lost lines of scientific inquiry.
• Central conclusion. The very difficult task of maximising innovation while minimising hazards to people and their environments is “ultimately a matter of political discourse”. It could be done more successfully if it embraced the twelve lessons.

Notes on the closing claims: - The objective is framed as a double one: maximise innovation and minimise hazards, not simply reduce harm. The report’s opening and closing say the same thing (Ch1, p. 11: “without stifling innovation or compromising science”). The Executive Director’s Preface uses the same pairing (“minimise environmental and health costs and maximise innovation”, p. 5) and makes the same innovation claim, more cautiously: “Many of the case studies suggest that wider use of the precautionary principle can help stimulate both innovation and science” (p. 4). - Placing the task in “political discourse” matches Ch16 (p. 183): choosing among precautionary responses is “an essentially political business”. - The only costs of over-precaution named are lost innovation and lost science (see the bias check). - The phrasing “proportionate and precautionary” treats proportionality as a counterweight to precaution. The EC Communication treats proportionality as one of the principles for applying precaution. So does the report itself: Ch1 (p. 13) lists “the proportionality principle” among the elements of the German Vorsorgeprinzip, and the asbestos chapter says the choice of precautionary measures “would depend on the proportionality principle” (Ch5, p. 60). This is a minor inconsistency, but it is internal to the report as well as with EU usage.

Case timeline#

Ch17 is not a case study. It does, however, invoke dated cases, and the table below sets those dates against the report’s own case tables to show the lags between warning and action behind the examples in Table 17.1.

Overall lag claims (from Ch16, which Ch17 summarises). The gap between identifying a problem and effective action was “many years or decades, and in some cases over a century”. Clear precautionary action remained scarce even after the principle was articulated in the 1970s–80s (Ch16, p. 168).

The authors’ own lessons and conclusions#

Kept distinct by type:

A. Historical claims the authors derive from the case evidence - Regulatory inaction led to costly consequences, some unforeseen and some unforeseeable (p. 192). - Early and even “loud and late” warnings were clearly ignored (p. 192). - Hazard appraisal was often too narrow (p. 192). - Regulatory actions were sometimes taken without considering alternatives or the conditions for real-world implementation (p. 192). - Most cases involved costly impacts on both health and environment (p. 193). - Past practice gave values little weight and demanded unequivocal proof before action (p. 193).

B. Conceptual contributions - A broad definition of the precautionary principle as a framework of foresight (p. 192). - The six-term typology of Table 17.1: risk, uncertainty and ignorance matched with prevention, precautionary prevention and precaution (p. 192). - The levels-of-proof framing: the evidentiary threshold distributes the costs of error and is political and ethical, and should be calibrated to harm, benefits, alternatives and error costs in both directions (p. 193).

C. Recommendations and advocacy - Fair, transparent and accountable procedures, as “good governance” to restore confidence (pp. 192–193). - Integrate health and environment in science and policy (p. 193). - Widen appraisal to social issues; involve stakeholders from the earliest stage (p. 193). - Recognise openly that controversies turn on values (p. 193). - Humbler, complexity-aware science (p. 193). - Explicit agreement on levels of proof (p. 193). - The twelve lessons (pp. 193–194).

D. Predictions and expected effects (forward-looking, testable in principle) - Society “may do substantially better” at balancing innovation and hazard (p. 192). - Stakeholder involvement may improve trust without necessarily stifling innovation or compromising science (p. 193). - The lessons would increase the chances of anticipating costly impacts, improve the balance of pros and cons, and minimise the costs of surprises (p. 194). - The precautionary principle “can also” bring benefits beyond harm reduction by stimulating more innovation, through diversity and flexibility, and better science (p. 194). - Balancing innovation and hazard “could be more successful” if it embraced the lessons (p. 194).

E. Self-limiting statements - Some harms could not have been foreseen (p. 192). - The lessons cannot remove dilemmas, eliminate uncertainty or avoid the consequences of ignorance (p. 194). - Over-precaution can be expensive (p. 194). - The task is ultimately political (p. 194).

Mechanisms and dynamics#

Only mechanisms that Ch17 states or directly implies are listed. Where Ch16 supplies the evidence it is cited, since Ch17 is its summary.

1. Treating ignorance as if it were risk. Folding every state of limited knowledge into “uncertainty” (p. 192) lets appraisal proceed as if all outcomes were known, which neglects surprise. Ch16 (p. 169): every activity in the cases went through some risk assessment. What was neglected was the “virtual certainty” that some factors lay outside its scope.

2. Knowledge states change over time, and long latency locks in harm. Table 17.1 dates the categories (p. 192). The Preface (p. 3) adds the dynamic that makes delay costly: long latency sets up decades-long “pipelines” of harm that cannot be stopped before anyone acts. The asbestos chapter names a related trap, the “latency lacuna” (Ch5, pp. 55, 60). Under changing conditions the risks of today’s exposures cannot be known for decades, so absence of evidence is taken as evidence of absence.

3. Narrow framing and disciplinary capture. Specialisation separated health from environment (p. 193). Appraisal omitted social issues (p. 193), alternatives and real-world implementation (p. 192). Lessons 3–5 describe blind spots inside the dominant discipline (lesson 3, p. 173), capture of appraisal by one discipline (lesson 4, p. 174) and gaps between assumed and real-world conditions (lesson 5, pp. 174–175). Ch16 (p. 174) says disciplinary capture “can lead to a form of ‘institutional’ ignorance”.

4. Standard and burden of proof as a way of distributing costs. Choosing a level of proof decides who bears the cost of error (p. 193). A historical demand for “unequivocal scientific proof” before action (p. 193) worked as a mechanism of delay. Ch16 (pp. 183–184) locates part of this in statistical convention, which prefers false negatives to false positives.

5. Values, legitimacy and trust. Controversies are driven by values and uncertainties (p. 193). Procedures that are not fair, transparent and accountable erode public confidence (pp. 192–193). Early inclusion is presented as a source of both information and legitimacy (p. 193).

6. Technological monopoly, lock-in and the scale of surprise. “Monopolies” such as asbestos and CFCs (p. 192) turn an unforeseen harm into a large one. Diversity and adaptability reduce the cost of surprises (lesson 7, p. 194). Ch16 adds the lock-in mechanics: - market prices that leave out environmental and health costs gave asbestos, halocarbons and PCBs an “unjustifiable advantage” and kept superior substitutes out of the market (pp. 176–177); - once a technology is committed, institutional and market processes reinforce it even if it is inferior (p. 177); - systems lock in early, for “arbitrary” reasons such as “chance and first-leader advantage” (pp. 186–187).

7. Interested parties and dependence on their information. Lesson 10 (p. 194) presupposes undue influence. Ch16 (pp. 178–180) documents it: - long delays despite early evidence on benzene (1897), asbestos (1898) and PCBs (1899); - a UK BSE regulator responsible first to industry; - the temporary lifting of the US ban on DES as a growth promoter in 1974 after “strong pressure from the farming lobby”, despite available alternatives; - poorly substantiated “refutations” of findings on PCBs and acid rain; - suppression of data and intimidation of publishers (Great Lakes); - attempts to discredit critics (BSE, asbestos); - dismissal of agency scientists (Californian sardines); - self-censorship by advisory committees (Southwood, 1988); - heavy reliance on risk information produced and owned by the firms being assessed.

Ch16 hedges one part of this: “Not all of these cases demonstrate the delaying or distorting effect of non-independent sources” of knowledge (p. 179). Its remedy is independent public information institutions (pp. 179–180). Ch17 itself only gestures at this mechanism.

8. Institutional obstacles to learning. Lesson 11 (p. 194). Ch16 (pp. 180–181) lists: - short government and business cycles; - changes of administration (UK rendering standards withdrawn in 1979; Californian sardine conservation reversed); - tension between levels of government and between departments (MAFF did not tell the Department of Health about BSE for about 17 months); - regulators defending their own past decisions; - different national readings of the same evidence (the US but not the UK banned scrapie-affected animals from food in the 1970s).

9. Complexity outrunning linear science. Feedback, synergy, thresholds and instability defeat reductionist appraisal (p. 193).

10. Paralysis by analysis. Demands for more information can become a way of delaying action (lesson 12, p. 194). Ch16 (p. 181) defines it as failure of timely hazard reduction caused by “either information overload or lack of political will”, and gives two examples: the US Supreme Court decision on benzene, which imposed “layer upon layer” of information requirements, and the Swann Committee’s 1969 warning that the call for more research should not hold up its recommendations.

11. Experts’ mental models: overconfidence and “hubris”. Ch17 names the attitude, calling for humility and attention to what is not known (p. 193). The Preface states the general claim more strongly: “Misplaced ‘certainty’ about the absence of harm played a key role in delaying preventive actions in most of the case studies” (p. 4). That is the Executive Director’s summary judgement; Ch17 does not repeat or demonstrate it. Ch16 supplies the concrete blind spots:

    • “no evidence of harm” read as “evidence of no harm”, when no evidence was being sought (BSE, p. 172);
    • the recurring belief that new practices had solved old problems (asbestos, from 1906, p. 173);
    • satellite data showing ozone depletion set aside as suspect (p. 173).

12. Regulation that creates new hazards. Actions taken without looking at alternatives or implementation conditions (p. 192). Examples: MTBE, tall stacks (Ch16, pp. 174–176), and substitutes that shared the original’s hazardous properties, such as fibrous replacements for asbestos and second-generation CFCs (Ch16, pp. 173–174, 177).

13. Distribution across groups and over time. The costs of error are distributed (p. 193). Ch16 (p. 182) adds that consensus is harder when pros and cons fall unevenly across groups or periods. The Preface (p. 3) names an asymmetry that tilts decisions towards inaction: the costs of preventive action are “usually tangible, clearly allocated and often short term”, while the costs of failing to act are “less tangible, less clearly distributed and usually longer term”.

14. Precaution redirecting innovation. Precaution can channel innovation rather than stop it (p. 194). Ch16 (p. 182) cites asbestos, halocarbons, PCBs and antimicrobials, and says curtailment “may” also give a competitive edge to the countries that lead such innovations. Ch16 also concedes that “in some cases it will be necessary to severely curtail or end innovation in a particular field” where society judges the risks unacceptable (p. 182).

Framing and language. - The report’s title pairing “late lessons / early warnings”, and the phrase “loud and late”, set up a narrative of neglected foresight. - “Surprises”, “monopolies”, “blind spots” and “paralysis by analysis” are the working metaphors. - “Humility and hubris” moralise epistemic stance. - “Balance” recurs, positioning the text as moderate. - “Maximising innovation whilst minimising hazards” echoes the language of innovation policy (whether deliberately cannot be told from the text; the same pairing is in the Preface, p. 5). - The chapter does not analyse the reassuring rhetoric used by proponents; Ch15’s BSE subtitle, “how reassurances undermined precaution”, signals that analysis is elsewhere.

Transferable insights (technology-neutral)#

Strength ratings apply to the insight as supported by this section and the report evidence it summarises.

1. Separate risk, uncertainty and ignorance, and match the response to the state of knowledge. Applying tools built for known probabilities to situations where the outcomes themselves are unknown produces false confidence. Knowledge states are also time-stamped: a hazard can move from ignorance to risk within a decade. Evidence: p. 192 (Table 17.1); lesson 1 (p. 193); Ch16 Box 16.1 (p. 170). Strength: strong as a conceptual distinction, long established in decision theory and adopted widely since. Moderate as a causal claim that misclassification drove the historical failures, which rests on case narratives rather than systematic comparison.

2. One can act under ignorance through strategies that do not depend on knowing the specific harm. Screen on intrinsic properties that predict hard-to-reverse harm (persistence, bioaccumulation, novelty, ready dispersal), weigh the irreversibility and scale of a commitment even when its consequences are unknown, monitor broadly over the long term, and keep options diverse and adaptable so that any surprise is smaller. Evidence: p. 192 (Table 17.1, “precaution” row); lessons 2 and 7 (pp. 193–194); Ch16 (pp. 170–171, 187). Strength: moderate. The logic is clear and the persistence argument is well illustrated (PCBs, CFCs, MTBE). How well these strategies reduce the size of surprises is inferred, not demonstrated.

3. The evidentiary threshold for action is a political and distributive choice. Setting the standard of proof decides who bears the cost of error. It should be calibrated to the severity and reversibility of the potential harm, the benefits claimed, the alternatives available, and error costs in both directions. Evidence: p. 193; Ch16 Table 16.1 (p. 184) and the Type I/II discussion (pp. 183–184). Strength: strong as an analytic point, well grounded in statistics and decision theory. Its application to particular choices is not worked out.

4. Harm is often preceded by warnings that are available but go unused. The failure is frequently less a lack of information than its non-use, including “loud and late” warnings. Evidence: p. 192; Ch16 (pp. 168, 178–179). Strength: moderate to strong within the case set (well documented for asbestos, benzene, PCBs, BSE and antimicrobials). Not generalisable to a base rate, because the cases were selected as false negatives (Ch1, p. 12).

5. Interventions themselves can create hazards when adopted without appraising alternatives and real-world conditions. Solutions to one problem that shift or create another, and substitutes that share the original hazard’s properties, are recurring patterns. Evidence: p. 192 (observation 2); lessons 5 and 7 (pp. 193–194); Ch16 (pp. 173–177). Strength: moderate, with several clear illustrations (MTBE, tall stacks, fibrous asbestos substitutes, second-generation CFCs).

6. Sectoral and disciplinary silos produce blind spots. Hazards that cross domains (health and environment, animal and human, air and water) fall between regulatory specialisms. Evidence: p. 193; lessons 3–4 (p. 193); Ch16 (pp. 173–174). Strength: moderate, illustrated across at least six cases.

7. Claimed benefits deserve the same scrutiny as risks. Unexamined claims of benefit or efficacy can keep a harmful practice going. Evidence: lesson 6 (p. 193); Ch16 (pp. 175–176): 1953 trial data showed DES ineffective (and “positively harmful”) in preventing miscarriage, yet it was not banned for another 20–30 years; the radiation “justification principle”. Strength: moderate. One compelling case (DES) plus general reasoning. Ch16 notes that industry also asks for benefits to be weighed, so the lesson cuts both ways.

8. Near-monopoly technologies magnify the scale of any surprise; diversity works as insurance. Evidence: p. 192 (Table 17.1); lesson 7 (p. 194); Ch16 (p. 187). Strength: suggestive. Sound in logic and illustrated by asbestos, CFCs and PCBs, but with no counterfactual evidence, and the trade-offs of diversity (economies of scale, standardisation, familiarity) go unexamined.

9. Timely action needs regulatory independence from interested parties, including independent sources of risk information. Evidence: lesson 10 (p. 194); Ch16 (pp. 178–180). Strength: moderate to strong in the documented cases (asbestos, benzene, BSE, DES). Ch17 itself only asserts it, and the tension with inclusiveness is acknowledged but not resolved. Ch16 itself notes that not every case shows non-independent information causing the delay (p. 179).

10. Institutions have their own obstacles to learning. Short time horizons, changes of administration, departmental silos, defence of past decisions and differing national cultures all delay action even when evidence is shared. Evidence: lesson 11 (p. 194); Ch16 (pp. 180–181). Strength: moderate, with well-chosen illustrations but no systematic analysis.

11. Workers, users, residents and other practitioners often spot emerging harm, or gaps between assumed and actual practice, before formal expertise does. Their knowledge should be sought and scrutinised as rigorously as expert knowledge. Evidence: lesson 8 (p. 194); Ch16 (pp. 177–178): asbestos and PCB workers, Love Canal, BSE slaughterhouse practice, Swedish farmers; and the counter-example of the “pensioners’ party fallacy”. Strength: moderate.

12. Controversies over new technologies often turn on values and purposes as well as facts. Treating them as purely technical, or demanding unequivocal proof before any action, worsens governance. Early and broad involvement adds information and may build trust. Evidence: p. 193; lesson 9 (p. 194); Ch16 (pp. 185–186, 188). Strength: suggestive / asserted as far as this report goes. The examples (GMOs, Brent Spar) are not case studies, the trust benefit is hedged with “may”, and the argument draws on outside science-studies literature.

13. Calls for more research can work as a mechanism of delay, so a trigger for action short of proof is needed. Evidence: lesson 12 (p. 194); Ch16 (p. 181). Strength: moderate. The report itself notes the counter-case (Great Lakes: widening the research increased uncertainty) and says the line between prudence and paralysis depends on who gains and who loses (Ch16, pp. 181–182).

14. Complex systems (feedback, thresholds, synergy) defeat linear, reductionist appraisal. Explicit attention to what is not known is a scientific virtue, not anti-science. Evidence: p. 193; Ch16 (pp. 184–185). Strength: moderate as a general claim (well supported in systems science). In this section it is asserted, not demonstrated.

15. Better information does not dissolve the dilemma. Decisions under high-stakes uncertainty stay political and value-laden, and even the best processes will be surprised. Evidence: p. 194. Strength: strong as an epistemic limit.

16. Precaution may redirect innovation rather than suppress it. Evidence: p. 194; Ch16 (p. 182); Preface (p. 4). Strength: asserted / suggestive. Anecdotal examples only. Ch16 itself concedes that some innovation pathways may have to be severely curtailed or ended (p. 182). The later empirical literature on regulation and innovation is mixed (see below).

17. Error runs in both directions, and over-precaution also has costs. Evidence: pp. 192–194; Ch5 (p. 60). Strength: asserted. The costs of over-precaution are named but not evidenced systematically anywhere in the report, which contains no false-positive cases (Ch1, pp. 12–13). The nearest thing is the hormones chapter, which records trade sanctions and calls the original EU ban a “political risk assessment” but judges its continuation probably justified (Ch14, pp. 153–154; see limitation 2). Ch1 (p. 16) nevertheless calls the risk of false positives “smaller but commonly feared”, an asymmetry the report asserts but does not show.

Limitations, contestation and bias check#

1. Genre: synthesis and advocacy, not analysis. - Ch17 is three pages. It cites almost nothing (only the EC Communication) and presents no evidence of its own. - The claim that the cases “both support and illustrate the need for” the lessons (p. 193) is not demonstrated in Ch17; the demonstration is in Ch16. - By the report’s own account, the lessons were “distilled” through editorial discussion (p. 192; Preface, p. 3) inside a pre-existing framework (ESTO) whose points the case material was used to “test or elaborate” (Ch16, p. 168). That is partly top-down rather than purely inductive, which raises a risk that the cases were read to confirm the framework.

2. Case selection: only false negatives. - All 14 cases are false negatives, agents once considered harmless that later proved harmful (Ch1, p. 12). - They were chosen because “sufficient is now known” about their impacts (Ch1, p. 11), which amounts to selecting on the outcome. - The editors tried to include false positives (unnecessary precautionary actions), inviting industry representatives to submit them, but “no suitable examples emerged”. Some 25 candidates in Facts versus fears (Lieberman and Kwon, 1998) turned out, on closer examination, “not to be robust enough for those who recommended them” to accept the invitation to put forward the strongest half-dozen. Possible candidates still mentioned are the ban on dumping sewage sludge in the North Sea and the “Y2K millennium bug” (Ch1, pp. 12–13). - So the set cannot estimate how often warnings turn out to be false alarms, or what precaution costs. The balance claims in Ch17 (pp. 192, 194) rest on an asymmetric evidence base. The editors were open about this in Ch1, but Ch17 does not repeat it. Ch1 (p. 16) goes further than the evidence by describing the risk of false positives as “smaller but commonly feared”. - Costs and benefits were not analysed. Ch16 (p. 168) says the question of the costs and benefits of actions and inactions “proved to be the most difficult question” for the case authors, and that “a general analysis lay beyond the scope of the current publication”. Ch17’s language of “balancing” costs of restriction against costs of permissiveness (p. 192) therefore has no systematic cost accounting behind it. - A partial counterweight inside the report that Ch17 ignores: hormones as growth promoters (Ch14, Bridges and Bridges). This is the one case in which a precautionary restriction and its costs are examined, and the authors are not uncritical of it. - They say the EU bans of 1985 and 1988 were taken “principally in response to public concern” and were not supported at the time by the EU’s own Lamming Committee or by WHO/FAO’s JECFA. Because neither committee was asked to characterise its uncertainty, “the original EU ban was, in reality, a political risk assessment” (p. 154). - They report “no good evidence” so far that the ban has protected public health (p. 153), and EU exports facing sanctions of about EUR 160 million a year (p. 153). - Yet they conclude that “more recent scientific research … probably justifies” continuing the ban (p. 154), and they draw lessons that fit Ch17: scientific committees should be asked to identify their uncertainties, and “rigorous and transparent mechanisms” are needed for weighing risks against benefits (p. 154). - Standpoint: Jim Bridges had been a scientific adviser to the EU at the WTO hearing on the ban (biographies, p. 195), so the criticism comes from someone on the side of the ban. - Relevance: the case bears directly on three Ch17 claims: that public values improve decisions (p. 193), that the level of proof should be explicit (p. 193), and that over-precaution has costs (p. 194). Ch17 draws on none of it.

3. Hindsight. - Ch1 (p. 11) says judgements should rest on “the spirit of the times”, not on the luxury of hindsight, and Ch17 concedes that some consequences were unforeseeable (p. 192). - Still, calling warnings “clearly ignored” is inevitably done knowing which warnings proved right. - The “ignorance” category is applied in retrospect: whether pre-1974 CFCs were truly unknowable is contestable (see the analysis of Table 17.1).

4. Standpoint of authors and editors. - Ch1 (p. 12) discloses that the case authors are “not without strong views” and were mostly participants in the histories they describe. - The editorial team includes a former Friends of the Earth director and leading science-studies scholars with published positions on precaution. Two editors co-wrote case chapters that Ch16 draws on heavily: Gee (asbestos) and Harremoës (MTBE). Two of the three TBT authors were Greenpeace Research Laboratories scientists (biographies pp. 197–198). The team also included consultants who had worked for industry (MacGarvin, Keys), and one case author (Lambert) consulted for the nuclear industry as well as environmental groups (p. 197). - In fairness, Ch1 raises the authors’ involvement itself (“brought to the attention of readers”) and says they “were expected to be as objective as possible” (p. 12). - This is not a reason to dismiss the conclusions, but it is a reason to look for independent corroboration, especially for the parts that argue from theory (inclusion yields better science; precaution fosters innovation; participation builds trust).

5. Diagnosis and remedy do not quite match. - Ch16 (p. 168) gives a two-part diagnosis. In “many” cases, adequate information existed but was not brought to decision-makers early enough or was “discounted for one reason or another”. In “some” cases (asbestos, PCBs, the Great Lakes, sulphur dioxide) warnings were “effectively ignored … because of short-term economic and political interactions”. Ch16 also documents undue influence and suppression (pp. 178–180). - The Preface goes further. It judges that “the absence of political will to take action … seems to be an even more important factor in these histories than is the availability of trusted information”, and says the question of why warnings were ignored is one “we largely leave to the reader” (p. 4). - Yet Ch17 describes most lessons as improvements to information (p. 194). That fits the “many” cases of information not reaching decision-makers or being discounted. It fits less well the cases, and the Preface’s judgement, where interests and political will drove delay; there, better information may not be enough. - Lessons 10–11 address this only partly. The framing fits the EEA’s mandate as an information agency, and a lens built on this report should give the political-economy mechanisms more weight than Ch17’s summary does. The report’s own Preface supports that weighting.

6. Vagueness and operational difficulty. - Terms such as “reasonable grounds for concern”, “adequate”, “full account” and “robust, diverse and adaptable” have no criteria attached. - Nothing says when there is enough knowledge, who sets the level of proof, or how inclusion and independence are to be reconciled. - Defining the principle as a “framework of thinking” (p. 192) makes it flexible, but hard to test and hard to hold to account.

7. Trade-offs between risks are under-engaged. - The only costs of over-precaution named are lost innovation and lost lines of inquiry (p. 194). Two others go unmentioned: substitute hazards, and benefits forgone (for example, the health benefits of a restricted product). - This is so even though the report’s own cases contain substitution hazards (asbestos substitutes, second-generation CFCs, and MTBE introduced partly in the drive to phase out lead; Ch16, pp. 173–177). The report frames these as failures to evaluate alternatives rather than as risk-for-risk trade-offs that precaution can itself create. - The standard critiques make exactly this argument: Graham and Wiener, Risk vs. Risk (Harvard UP, 1995), and Sunstein, Laws of Fear: Beyond the Precautionary Principle (Cambridge UP, 2005), the latter arguing that strong versions of precaution are incoherent because risks lie on all sides. Ch17’s “costs of being wrong in both directions” (p. 193) partly answers this but does not engage it.

8. The innovation and science claims are asserted, not shown. - “Stimulating both more innovation … and better science” (p. 194) is supported in Ch16 (p. 182) only by examples in which curtailment was followed by substitution. There is no systematic evidence. - The later literature on regulation-induced innovation (the “Porter hypothesis”) remains mixed. Ambec, Cohen, Elgie and Lanoie, Review of Environmental Economics and Policy 7(1):2–22 (2013), is a twenty-year review; the citation was verified, but only the RePEc summary of the abstract was seen, so read it directly before relying on it.

9. The trust claim is hedged and unevidenced (p. 193).

10. Brent Spar cuts both ways. - The report treats “oil-rig disposal” as a controversy about values (p. 193; Ch16, p. 188). - Secondary accounts (a Wikipedia summary, not verified against primary sources) record three things. In September 1995 Greenpeace apologised to Shell for its estimate that the installation held over 5,500 tonnes of oil. A Det Norske Veritas audit in October 1995 endorsed Shell’s much lower figures. In 1998 OSPAR decided that disused offshore installations should in general not be dumped at sea, with derogations for some very large structures. (The report itself names neither Shell nor Greenpeace in this context; see pp. 188, 193.) - The episode supports the chapter’s point that values drive such controversies. It also shows that value-driven outcomes can rest on disputed factual claims, which complicates any implication that taking in public values straightforwardly improves decisions. - Primary sources to check: Greenpeace’s September 1995 letter to Shell; the DNV audit (October 1995); the UK Natural Environment Research Council’s 1996 report on disposal options; OSPAR Decision 98/3.

11. The military intelligence analogy is unsupported and double-edged (p. 193). - No source is given. - Intelligence history also contains costly false positives produced by precautionary threat assessment. The pre-2003 judgements on Iraqi weapons of mass destruction are the obvious example. The 2004 US Senate Select Committee on Intelligence report found most key judgements in the 2002 National Intelligence Estimate overstated or not supported by the underlying intelligence. The 2004 UK Butler Review was more guarded: it found that the September 2002 dossier went to the outer limits of the available intelligence and that caveats had been dropped. (Auditor’s summary from general knowledge, not re-verified here; check the reports before citing.) - That is outside the report and after it. It suggests the analogy shows why error costs are two-sided, not that precaution is inherently right.

12. Loose ends in the editing. - Ch16 twice refers readers to an “‘Implications’ part of this report” (p. 184): once for elaboration of Table 16.1’s levels of proof, and once for the bias towards false negatives. No part with that name exists: the Contents (p. 10) go from Ch17 straight to the author biographies. The Preface (p. 3) says implications will be explored in a separate EEA publication. - Box 16.1 (p. 170) points instead to the elaboration of levels of proof “in the last part of this report”, which can only be Ch17. Ch17 does carry a levels-of-proof paragraph (p. 193), but it does not take up the Type I/Type II bias. The likeliest reading is that Ch17 is a compressed version of a planned “Implications” part, which could explain how thin the levels-of-proof discussion is. This is inference; the report does not say so. - Lesson 10’s wording (“independence of interested parties”) is ambiguous (see §8 above).

13. Fairness in the other direction. - Ch17 is more measured than the caricature of precautionary advocacy. It concedes that some harms were unforeseeable, that over-precaution has costs, and that the lessons cannot resolve dilemmas. - It frames error costs symmetrically, calls for scrutiny of benefits (something industry also asks for), and ties the level of proof to proportional considerations rather than to any suspicion at all. - Elsewhere the report is candid in ways Ch17 compresses: that restriction should stop when research shows a concern unfounded (Ch16, p. 173); that a better false-positive/false-negative balance would impose some costs of restricting things later shown safe (Ch5, p. 60); that participation has limits (Ch16, p. 188); and that political will may matter more than information (Preface, p. 4). - Critics who equate the principle with “act on any suspicion” or “ban until proven safe” are not describing this text. - Its account of national differences (Ch16, p. 168: variation within institutions on both sides of the Atlantic rather than a simple EU-versus-US divide) anticipates the later comparative study by Wiener, Rogers, Hammitt and Sand (eds), The Reality of Precaution: Comparing Risk Regulation in the United States and Europe (RFF Press, 2011). My reading, not checked against the book in this session, is that the study argues neither side is uniformly more precautionary.

Later evidence relevant to the headline claims (outside the report; verification status given) - False positives. Hansen, Krayer von Krauss and Tickner, “Categorizing mistaken false positives in regulation of human and environmental health”, Risk Analysis 27:255–269 (2007). - They identified 88 cases claimed as false positives and judged only four to be authentic regulatory false positives: Southern Corn Leaf Blight, swine flu, saccharin, and food irradiation with respect to consumer health. (Verified from the abstract via Europe PMC.) - Caveats: the authors were close to the 2001 report (Krayer von Krauss co-wrote its MTBE chapter with Harremoës, p. 8, and Tickner is listed among its peer reviewers, p. 6), and the result depends on their definition of a false positive. - The 2013 EEA volume revisits false alarms; that is covered in another strand. - Antibiotic growth promoters. Article 11(2) of Regulation (EC) No 1831/2003 allowed antibiotics other than coccidiostats and histomonostats to be marketed and used as feed additives only until 31 December 2005 (verified on EUR-Lex). - The industry’s challenge to the 1998 ban (Table 9.2, p. 99, says a judgement was expected at the end of 2001) was Case T-13/99, Pfizer Animal Health v Council. My recollection is that the Court of First Instance upheld the ban in September 2002; this was not verified here, as the EUR-Lex text could not be retrieved. - The typology. Stirling later elaborated risk, uncertainty, ambiguity and ignorance (for example, “Keep it complex”, Nature 468:1029–1031, 2010; citation verified). That fills the gap left by Table 17.1’s omission of ambiguity. - Property-based screening. In-report precedents: the Third North Sea Conference (1990) on “persistent, toxic, and liable to bioaccumulate” substances and the Stockholm Convention on POPs (2001), both in Table 1.2 (p. 14), and Swedish chemicals policy (Ch16, p. 183). After the report, EU chemicals regulation (REACH, Regulation (EC) No 1907/2006, Annex XIII) set PBT/vPvB criteria. That last point is from the auditor’s general knowledge, not verified in this session.

Notable quotes#

Short phrases only, each under 15 words. The full text is in the extract.

1. “loud and late” (warnings that were clearly ignored), p. 192
2. “an overarching framework of thinking that governs the use of foresight”, p. 192
3. “fewer technological ‘monopolies’ such as asbestos and CFCs”, p. 192 (Table 17.1)
4. “more humility and less hubris”, p. 193
5. “‘what we don’t know’ as well as on ‘what we do know’”, p. 193
6. “the size, nature and distribution of the costs of being wrong”, p. 193
7. “a key political decision with profound ethical implications”, p. 193
8. “Avoid ‘paralysis by analysis’”, p. 194 (lesson 12)
9. “over-precaution can also be expensive”, p. 194
10. “ultimately a matter of political discourse”, p. 194

Open questions#

1. Who drafted Ch17, and were there disagreements within the editorial team or the EEA Scientific Committee about the lessons or their wording? The report records none.
2. How far were the twelve lessons derived from the cases, and how far imported from the ESTO framework (Ch16, p. 168)? Would an independent reading of the 14 cases produce the same list?
3. Did the “separate EEA publication” on implications (Preface, p. 3) appear, and did it develop the levels-of-proof argument that Ch16 promises (p. 184)?
4. Can Table 17.1’s categories be applied in real time rather than retrospectively? How would one tell “institutional” from “societal” ignorance (Ch16, p. 171) before a surprise? The report treats pre-1974 CFCs as a strong case of ignorance (Ch16, p. 169). Is that right, given the 1907, 1934 and 1973 evidence (Table 7.1), or was the problem partly that the components were never assembled? And where does the unclassified 1959–65 period for mesothelioma fall?
5. Is there empirical evidence that screening on persistence and bioaccumulation, or technological diversity, actually reduces the scale or cost of later surprises?
6. After 2001, did any regulatory regime make its choice of level of proof explicit and case-specific, as p. 193 recommends, and with what effect on the distribution of error costs?
7. Did early stakeholder involvement improve trust without stifling innovation (p. 193)? What does post-2001 evidence from participatory exercises show?
8. How should the tension between lesson 10 (independence) and lessons 8–9 (inclusion) be resolved in practice, particularly when the parties with the most information are also the most interested?
9. What is the base rate of false alarms relative to false negatives? Hansen et al. (2007) is one answer; how robust is it to other definitions?
10. How does the goal of “maximising innovation whilst minimising hazards” deal with innovations whose benefits are themselves reductions in hazard (risk-for-risk cases like MTBE and lead)?
11. If delay is driven largely by interests and lack of political will (Ch16, p. 168, for some cases; Preface, p. 4, more broadly), how much can lessons that are mostly about information (p. 194) achieve without changes to incentives, liability or the allocation of the burden of proof? Ch16 touches on these tools (polluter pays, taxes, liability regimes, internalising costs, pp. 176–177), but Ch17 does not carry them forward.
12. How should the hormones case (Ch14, pp. 153–154) be read against Ch17: a precautionary restriction driven by public concern, not backed by expert committees at the time, with measurable trade costs and no demonstrated health benefit so far, yet judged “probably” justified by later research? Would Ch17’s levels-of-proof framework (p. 193) have handled it better?

Audit log#

Independent audit against the extract (all of pp. 192–194, which was also re-rendered from the PDF) and against the cited pages of the same PDF (Preface pp. 3–5, Acknowledgements p. 6, Contents pp. 7–10, Ch1 pp. 11–16, Ch5 pp. 55, 58, 60–61, Ch7 p. 83, Ch9 pp. 94, 98–99, Ch16 pp. 168–191, biographies pp. 195–199). All Ch17 quotations were checked verbatim, and the twelve lessons were confirmed identical in Ch16 and Ch17. The core summary of Ch17 was accurate; the changes below correct cross-references, rebalance several interpretive claims and add omitted context. No web sources were consulted in this audit.

• “What was read”: added that the bold action labels in Table 17.1 are also lost in extraction.
• Authorship: added that the Preface is signed by EEA Executive Director Domingo Jiménez Beltrán, and listed the Scientific Committee members (p. 6).
• Harremoës: added that he co-wrote the MTBE chapter (Ch11) with Krayer von Krauss, which matters because MTBE is a heavily used example.
• Stirling/ESTO: added that Ch16’s references record Stirling (1999) ESTO as cited in the EC Communication, so one framework fed both.
• Evident stance: softened “shaped the conclusions as much as the case evidence did” to a claim the text supports.
• Political context: flagged that the months for COM(2000)1 and Nice come from outside the report, and gave the report’s own name for the Nice text.
• §1 definition note: corrected the claimed contrast with Ch1. Ch1 (p. 13) is itself broad, and says precaution “requires much more than” a trigger.
• Table 17.1 analysis 1: noted that 1959–65 is left unclassified.
• Table analysis 2 (asbestos): added Ch5 p. 55 evidence of mesothelioma observations in the 1940s–50s and 1955, which qualifies “ignorance before 1959”.
• Table analysis 2 (antimicrobials): noted that the report dates Swann variously 1967, 1968 and 1969.
• Table analysis 2 (CFCs): rebalanced. Ch16 (p. 169) and Farman (p. 83) both treat pre-1974 CFCs as ignorance; the 1907/1934 work was general lab chemistry; the “institutional ignorance” reading was downgraded to a weaker, qualified objection.
• Table analysis 3: corrected the page for novelty, persistence and dispersal (pp. 170–171, not p. 171); added irreversibility and scale; added Ch16’s different remedies for institutional and societal ignorance; added in-report precedents for property-based triggers (North Sea 1990, Swedish policy).
• Table analysis 4: added that Box 16.1 and Ch16 p. 186 do address values; the gap is formalisation, not neglect.
• Table analysis 5: added Ch1 p. 12’s explicit statement about US “precautionary prevention”, and the medical roots of the term (p. 13).
• Procedural requirement: corrected the Box 16.1 cross-reference and quoted its wording.
• §4: flagged that the NRC, RCEP and Nice sources sit in §16.3, outside the case material; added Ch16’s caveats on participation (pp. 178, 188).
• §5: noted a possible alternative parse of the “demand for unequivocal proof” sentence; added Ch16 p. 183 on transatlantic contestation of EU GMO assessment.
• §6: corrected “originates with” the antimicrobials authors (the phrase is in Ch9 p. 98 and Ch1 p. 15; which came first is unknown); added the p. 178 public-values link; added Ch16’s prospective/retrospective reply (p. 181) and the p. 173 “wrong agent” caveat.
• §7: added the Box 16.1 distributive point and the asbestos chapter’s concession on false positives (Ch5, p. 60); refined “does not say who should choose” (it says the choice is political but gives no body, procedure or weighting).
• §8: added the qualifiers Ch16 attaches and Ch17 drops (participative context, proportionality to scale, no single set of prescriptions), and flagged that §16.3 goes beyond the case evidence.
• §9: added the Preface’s parallel and more cautious innovation claim (p. 4) and its minimise/maximise pairing (p. 5).
• Case timeline: antimicrobials row now includes the 1970s partial implementation of Swann, the 1997 Swedish report and the 2000 WHO recommendation, and qualifies the lag; CFC row adds the countries that followed the 1977 US ban and the marginalisation of the 1980 European restriction.
• Authors’ conclusions D: restored the modal “can” on the innovation claim.
• Mechanism 2: added the “latency lacuna” (Ch5, pp. 55, 60).
• Mechanism 3: made the lesson and page mapping precise (institutional ignorance is tied to disciplinary capture, p. 174).
• Mechanism 6: corrected “first-mover” to the report’s “first-leader advantage”.
• Mechanism 7: corrected to the temporary lifting of the US DES ban; added Ch16’s hedge that not every case shows distortion from non-independent information (p. 179).
• Mechanism 10: added Ch16’s definition of paralysis (information overload or lack of political will).
• Mechanism 13: added the Preface’s point about the asymmetric visibility of costs (p. 3).
• Mechanism 14: corrected the competitive-edge claim to “may”, and added Ch16’s concession that some innovation may need to be curtailed or ended.
• Insight 2: added irreversibility, scale and dispersal; corrected the pages.
• Insight 7: corrected the DES wording (not banned for another 20–30 years; also “positively harmful”).
• Insight 9: added the p. 179 hedge.
• Insights 16–17: added the Ch16 concession, and Ch1 p. 16’s unevidenced claim that false positives are the “smaller” risk.
• Limitation 2: corrected the Facts versus fears account (it was the recommenders who declined) and added the candidates Ch1 still mentions; added Ch1 p. 16; added Ch16 p. 168’s admission that no general cost-benefit analysis was done.
• Limitation 4: added the case-chapter authorship of two editors, the Greenpeace affiliation of two TBT authors, and the industry consultancy of some editors and authors, for balance.
• Limitation 5: corrected an overstatement. Ch16 attributes ignored warnings to economic and political interactions in “some” cases, not “often”, and names information failures in “many”; added the Preface’s stronger political-will judgement (p. 4).
• Limitation 11: corrected the Butler Review characterisation (it did not find judgements unsupported in the way the Senate report did); flagged as unverified general knowledge.
• Limitation 12: added Box 16.1’s pointer to “the last part of this report”; marked the Ch17-as-compressed-“Implications” reading as inference.
• Limitation 13: added the report’s candid concessions elsewhere that Ch17 compresses.
• Later evidence: specified Krayer von Krauss’s role (MTBE co-author); added in-report precedents for property-based screening and named REACH Annex XIII as unverified recollection.
• Open questions 4 and 11: updated to match the corrected readings.

Second audit (25 September 2026)#

Independent re-check of the notes and digest. Ch17 (pp. 192–194) was re-read in full from the extract and re-rendered from the PDF. Every cross-reference was checked against a fresh text extraction of the same PDF: the Preface and acknowledgements (pp. 2–6), Contents (pp. 7–10), Ch1 (pp. 11–16), Ch5 (pp. 55, 58, 60–61), Ch7 (p. 83), Ch9 (pp. 94, 98–99), Ch14 (pp. 153–154), Ch16 (pp. 168–191) and the biographies (pp. 195–199). All Ch17 quotations and the ten “Notable quotes” are verbatim. The twelve lessons are identical in Ch16 and Ch17. No factual errors were found in the summary of Ch17 itself. The changes below correct cross-references, soften overstatements and add omissions. One web check (Case T-13/99) failed to load, so that item stays marked unverified.

• Authors (Stirling): corrected the quote to “provided the initial framing of this analysis”; softened “one framework fed both” the lessons and the EC Communication to a shared lineage (the report does not say ESTO supplied the Communication’s trigger wording).
• Evident stance, §1 and limitation 1: made the “test or elaborate” wording exact (the case material tested “many of the points arising from the ESTO studies”); added Ch16’s counter-statements that the key issues “emerged from the case studies” (p. 168) and are “well grounded in the empirical detail” (p. 182).
• Table 17.1 analysis 2 (antimicrobials): changed “internally inconsistent” to “partly inconsistent”. Ch9 (p. 94) has the committee set up in 1968 and reporting in 1969, which reconciles “evidence available in 1968” with 1969; only “1967” (p. 172) and “early 1968 recommendation” (p. 174) conflict.
• Table 17.1 analysis 3: added Ch16’s own hedges on acting under ignorance (“not quite as simple as it seems”; “rather better than in the past”, p. 170).
• §3: made the “most of the cases” assessment more precise (a modest majority; asbestos is also chiefly a health case).
• §7: added that Ch17’s range of levels of proof omits the lowest rung in Box 16.1/Table 16.1 (“scientific suspicion of risk”), which makes the EC standard look like the floor.
• §9: added that the “proportionate and precautionary” pairing also conflicts with the report’s own usage (Ch1, p. 13; Ch5, p. 60), not only with the EC Communication.
• Case timeline (antimicrobials): updated the Swann-date note to match.
• Mechanism 11: added the Preface’s claim (p. 4) that misplaced certainty about absence of harm delayed action in most cases, labelled as the Executive Director’s judgement and not demonstrated in Ch17.
• Framing and language: softened “deliberately borrows” innovation-policy language to “echoes” (intent cannot be told from the text).
• Insight 17: qualified “not evidenced anywhere” and pointed to the hormones chapter as the nearest in-report evidence on the costs of a contested restriction.
• Limitation 2: added the omitted hormones case (Ch14, pp. 153–154): original EU ban called “a political risk assessment”, no good evidence yet of a health benefit, about EUR 160 million a year in sanctions, but continuation “probably” justified; author Jim Bridges had advised the EU at the WTO (p. 195).
• Limitation 4: added, for balance, Ch1’s disclosure that authors “were expected to be as objective as possible” (p. 12).
• Open questions: added question 12 on the hormones case.
• Digest: corrected the Swann dating; noted the omitted lowest level of proof; added misplaced certainty to the mechanisms; replaced “editorial team rooted in science and technology studies and NGO work” with an accurate description of the team (including its engineering and industry-consultancy members); added caveats on the hormones case and the proportionality inconsistency.