LL2-06 — Ch6 Beryllium’s ‘public relations problem’#

Late lessons from early warnings: science, precaution, innovation (EEA Report No 1/2013). Report pages 131–150 (PDF pages 133–152).

• Chapter: pp. 131–140, with Table 6.1 on p. 140 and references on pp. 141–144.
• Panel (Guidotti): pp. 145–150.

Reading note. I read the full extract in order, through the last page marker (PDF 152 / p. 150). I then checked it against the PDF visually for the summary box, pp. 133–135, pp. 137–140, Table 6.1 and all panel pages. The extraction is faithful. Where the PDF uses emphasis, it is noted below.

Authors and standpoint#

Chapter: David Michaels and Celeste Monforton.

• Michaels. At publication he was Assistant Secretary of Labor for OSHA. He wrote the chapter while on the George Washington University faculty, and it states personal views only (fn 1, p. 131).
• Earlier, as Assistant Secretary of Energy for Environment, Safety and Health, he “directed the DOE efforts to issue a stronger beryllium standard and develop a programme to provide compensation payments to workers with CBD” (p. 132).
• Monforton. The section does not state her position.
• Acknowledged: David Kriebel and Christina Morgan (fn 2, p. 131).

Sources and method. - A review of documents plus Michaels’s “personal knowledge” (p. 132). - Documents come from government files and from “attorneys who obtained them via litigation” (p. 132). - Many of the internal company documents cited are listed as “available from authors” (pp. 141–144). - The chapter is closely related to (a reworked version of) the authors’ 2008 Public Health Reports article of the same title (cited by the panel, p. 150). The two texts differ in places: the 2008 article’s online text (PMC2099330, checked 25 Sept 2026) has different section headings, and its lesson 2 lacks the words “for misinterpretation” that this chapter adds (p. 140). - Its organising concept, “manufacturing uncertainty”, is the authors’ own (Michaels and Monforton, 2005; Michaels, 2008) (p. 140).

Stance. - Participant-advocate. Michaels was the DOE official who issued the 1999 rule over industry opposition. - Critical of: - Brush Wellman (now Materion); - its PR and “product defence” contractors; - the DOE weapons offices and the Department of Defense; - the Reagan and G.W. Bush administrations’ handling of OSHA. - Credits: - the Atomic Energy Commission (AEC)’s early action (“To its credit”, p. 133); - the 1949 limit as “a great step forward” (p. 133); - the Brush–NIOSH partnership as “productive” (p. 138); - Brush-affiliated researchers’ later findings (p. 139). - No precaution language. The chapter never uses the word “precaution”. Its normative anchors are: - “The public health paradigm requires that the best available evidence be used to protect the public” (p. 140); - “prudent public health policy” (pp. 139, 140). - Framing of the case. It is framed less as acting under deep uncertainty and more as delay after the evidence was “sufficient” (by the early 1990s, p. 140). - Tension. At publication Michaels headed the agency whose inaction the chapter records (“2012 OSHA has yet to propose new workplace beryllium standard”, Table 6.1, p. 140). The disclaimer handles this formally only.

Summary box (p. 131). An EEA-style abstract covering both chapter and panel. It concedes the panel’s proposal “may be perceived as letting them ‘get away with it’”, but adds that “the end result may be better public policy and corporate responsibility”.

Panel: Tee L. Guidotti, “Corporate behaviour in the regulation of beryllium: could there have been a different outcome if the company had room to turn around?” (pp. 145–150).

• It is an adaptation of “a critical review prepared for the European Environmental Agency” (fn 3, p. 145).
• His institution is not stated. He cites his own work in occupational medicine and epidemiology (p. 150).
• Attitude to the chapter. He calls it “a comprehensive history” and “a valuable contribution” (p. 145). He declines to judge culpability and reframes the case as organisational behaviour and “pathway dependence” (p. 145).
• Partial dissent:
• defends the 1948–49 standard-setting (pp. 147–148);
• criticises the chapter’s “discounting” lesson (p. 148);
• qualifies the “absence of evidence” lesson for the BeLPT (p. 148);
• says reluctance “appears to have been driven mostly” by uncertainty over the blood test (p. 147), while also conceding that “much of the opposition was self-serving” (p. 147).
• Not an industry response. No company voice appears in its own terms. Guidotti: “Doubtless, the company has a different narrative” (p. 145).

Section-by-section notes#

Summary box (p. 131)#

• “Scores” of nuclear weapons workers have chronic beryllium disease (CBD), “a progressive and irreversible inflammatory lung disease”.
• Evidence grew that CBD occurred below the standard. Industry had “a strong financial incentive to challenge the data”, hired PR and “product defence” firms, and, when denial failed, adopted “a new rationale to delay”.
• Lessons flagged:
• hazards across “the entire product life cycle”;
• it is “unlikely that many secondary users and recyclers have the expertise, resources and knowledge necessary” to prevent disease;
• “manufacturing uncertainty” as a strategy “to prevent or delay regulation or victim compensation”.

6.1 Introduction (p. 132)#

• Richardson’s admission. In April 2000 Energy Secretary Bill Richardson acknowledged on national TV (ABC 20/20) that his department had “collaborated with the beryllium industry to defeat a 1975 attempt” by OSHA to cut exposures (the authors’ paraphrase, not Richardson’s words). This followed an exposé (Roe, 1999, The Blade).
• Quote: “Priority one was production of our nuclear weapons … [the] last priority was the safety and health of the workers”.
• The authors call such admissions of “deception that resulted in death and disability” rare and “long overdue”.
• DOE’s new level. DOE lowered “the level that triggered protection” from 2.0 to 0.2 μg/m³ (an action level; see limitations). Beryllium is “so hazardous that no safe level of exposure has ever been established”.
• Coverage gap. “The DOE standard covers only workers employed in the nuclear weapons complex”.
• OSHA. It has acknowledged its limit is inadequate but not updated it; its standard covers private-sector workers.
• Who is exposed. NIOSH estimated 28,000–107,000 exposed US private-sector workers, only 1,500 of them in primary production (Henneberger et al., 2004).
• Scope. US-focused. A UK investigation exists (Watterson, 2005). Brush had European factories and influenced policy “throughout the world”. The European dimension is not developed.

6.2 Early warnings and the first standard (pp. 132–133)#

• First harm. 1930s: “dozens” of Massachusetts fluorescent lamp workers developed acute beryllium disease (ABD) (Hardy, 1950). Workers “could not safely work with beryllium without respiratory protection” (p. 132).
• Weapons. The Manhattan Project and Cold War made beryllium “a vital component in nuclear weapons” because it raises “yield”. There were many early cases among plant workers “and among residents living near these facilities” (pp. 132–133).
• AEC response. “To its credit,” the AEC “acted quickly”. It funded studies and drew on scientists who had “virtually invented the field of radiation protection”. As the main buyer it “tacitly assumed responsibility” for research (p. 133).
• 1947 AEC report, Public Relations Problems in Connection with Occupational Diseases in the Beryllium Industry (Tumbelson, 1947):
• acknowledged “obvious moral responsibility” as well as PR problems, the latter worsened because some processing factories sat in “more populous areas”;
• warned indignation could reach the scale of “silicosis or radium poisoning”;
• noted recruitment problems “because of local prejudice … engendered by actual and rumored experience with beryllium poisoning” (p. 133). (My gloss, not the authors’: local lay knowledge was already acting as an economic signal.)
• The “taxicab standard” (p. 133):
• In 1948 Merril Eisenbud (AEC industrial hygienist) and Willard Machle (a physician consulting to the firm building Brookhaven Laboratory) picked 2 μg/m³ for workplaces and 0.01 μg/m³ for communities “in the back seat of a taxi”. They did so “in the absence of an epidemiological basis” (Eisenbud, 1991).
• Stokinger later called the method a “crude analogy” (1966).
• The AEC adopted it “tentatively” in 1949, reviewed it annually for seven years, then made it permanent and wrote it into supply contracts. OSHA adopted it in 1971.
• Egilman et al. (2003) suggest it was chosen on “feasibility not health protection”. The authors report this without adjudicating.
• Why it was accepted (p. 133). It was “very stringent for its time”. Two factors “probably” helped:
• the human cost of acute disease was so great that the accompanying “‘public relations problems’ threatened the AEC’s mission”;
• weapons work was “well funded — essentially a ‘cost-plus’ operation”, with the government reimbursing costs plus a percentage as profit. The implication (drawn by me, not spelled out by the authors) is that compliance costs could be passed on to the government.
• Brush Wellman (“now known as Materion”) was both vendor and contractor, running an AEC plant in Ohio from 1950 to 1956.
• Outcome. An “immediate success; ABD virtually disappeared”, and CBD appeared to fall. But questions soon arose about CBD at lower exposures (p. 133).

6.3.1 Evidence of CBD below 2 μg/m³ (pp. 133–135)#

• Registry evidence.
• Harriet Hardy founded the Beryllium Case Registry (BCR) at Massachusetts General Hospital in 1952, with AEC funding.
• By 1972 it held at least 20 CBD cases among workers who started after 1949 (NIOSH, 1972). By 1975 it held at least 36 (OSHA, 1975).
• The authors hedge the inference: this was “suggesting the disease might be occurring” below the limit (p. 134). Post-1949 start dates do not by themselves show below-limit exposure.
• Non-occupational cases: people who “laundered the clothes of workers, drove a milk delivery truck”, or “tended cemetery graves near a beryllium factory” (p. 134).
• Atypical dose–response (p. 134):
• CBD “did not follow the usual exposure-response model”.
• In 1951 Sterner and Eisenbud found exposure uncorrelated with severity and hypothesised immune susceptibility.
• Stokinger’s 1966 AEC-sponsored text: cases “from exposures to seemingly trivial concentrations … no dose-response relationship appears to hold”. The bold is the authors’ emphasis.
• A threshold “might not be possible to identify”.
• The conundrum (p. 134):
• Community cases were “viewed as anomalous, or the result of episodes of high exposure”.
• Systematic worker studies showed CBD falling, prompting speculation the limit was “overly conservative” (Breslin and Harris, 1959; Stokinger, 1966).
• The authors’ hedged explanation (“With hindsight, we can speculate”): immune-mediated disease, “considerable inter-individual variability in susceptibility”, and a dose–response “probably driven by ‘peak’ exposures — possibly of very short duration — which standard methods of exposure assessment do not detect”.
• Further signals (p. 134):
• In 1974 a delegation from NGK (a Japanese producer with a US plant) brought US industry executives a report of five CBD cases below 2 μg/m³ (Kohara, 1974; Shima, 1974). The chapter does not say what was done with it.
• The reference list shows the Kohara item is a letter from an NGK director to H.C. Piper, a Brush Wellman vice-president (9 August 1974), and the Shima item is a letter to a Cleveland Clinic physician (p. 142). So Brush itself received the report.
• Four cases occurred at one scrap reclamation facility consistently below the limit (Cullen et al., 1987). The reference title says “precious metal refinery” (p. 141).
• Detection (p. 135):
• CBD begins before it is visible on X-ray (Newman et al., 1996).
• The first CBD diagnoses by blood lymphocyte proliferation test (BeLPT) were reported in 1983 (MMWR).
• “By the end of the decade” (the 1980s; Kreiss et al., 1989; Mroz et al., 1991) improved techniques let clinicians “more easily identify” beryllium sensitisation (BeS), the precursor to CBD.
• Screening results (p. 135), across weapons, ceramics, primary production, machining and copper–beryllium alloys:
• CBD prevalence 0.1–4.4 %;
• BeS prevalence 0.9–9.9 %;
• 9–100 % of sensitised workers had CBD;
• cases occurred even among “clerical workers and security guards” with “bystander exposure”;
• BeS progresses to CBD at “6 % to 8 % per year”, though it is unknown whether all sensitised people progress (Newman et al., 2005).

6.3.2 Cancer (p. 135)#

• NIOSH, 1977: “probably no compounds known to man give so consistent a carcinogenic response in so many animal species”.
• Proposals. This “compelled OSHA to propose” 1 μg/m³ (8-hour time-weighted average, TWA), and NIOSH recommended 0.5 μg/m³.
• Chronology note: Table 6.1 dates the OSHA proposal to 1975, before the 1977 testimony. The reference list confirms the problem: NIOSH (1977) is the testimony of Deputy Director Edward J. Baier at the 19 August 1977 public hearing on the beryllium standard (p. 143), and OSHA’s proposal is in the 1975 Federal Register (p. 143).
• Brush’s response. It assembled “a team of toxicologists, statisticians and physicians” to challenge the proposal (Michaels, 2008). Its stated stake: once labelled a carcinogen, usage would shrink until “it would no longer be a viable industry” (Brush Wellman, 1977).
• Why the rule failed. “Intense lobbying by the industry and the US Departments of Defense and Energy and the election of President Ronald Reagan” (Roe, 1999).
• Later classification. NIOSH cancer epidemiology continued (Steenland and Ward, 1991; Ward et al., 1992).
• IARC: Group 1 in 1994, reaffirmed 2009.
• US National Toxicology Program (NTP): “known human carcinogen” in 2002.

6.3.3 The industry’s public relations efforts (pp. 135–136)#

• Motive. A “no safe level” or carcinogen designation would lead customers “to look for substitutes” (Brush Wellman, 1977; Hanes, 1992b). The industry decided “to be proactive in shaping the interpretation of scientific literature” (p. 135).
• Powers–Preuss memo, 1987 (retired executive-consultant and Corporate Medical Director) (pp. 135–136):
• Argued “the literature on Be published in the last twenty years has been very damaging”: it was cited by customers and by OSHA and EPA, which “cite these erroneous documents”.
• Proposed a textbook “financed by Brush (or Brush and NGK?)”, which “to be fully acceptable and credible … will have to be published under the auspices of some not-for-profit organisation such as a university or medical group”, plus papers “in prestigious medical books”.
• Result: Beryllium: Biomedical and Environmental Aspects (1991), edited by Powers, Preuss and “a respected academic physician” (Rossman et al., 1991).
• Caveat: the chapter documents the plan and the resulting co-edited book, but not who financed the book or whether it appeared under a not-for-profit’s auspices. The reference lists Williams and Wilkins, Baltimore, as publisher (p. 143).
• Hill & Knowlton (p. 136), a firm known for “manufacturing and promoting scientific uncertainty for the tobacco industry”:
• Its 1989 proposal, which the authors say “echoed” the 1947 AEC memo: “Beryllium undoubtedly continues to have a public relations problem”. Aim: to change “common erroneous attitudes” and “to dispel myths and misinformation” (Marder, 1989a).
• Offerings: “an authoritative white paper … the most definitive document available”; projects engaging outside scientists in independent reviews of Brush materials (the authors’ paraphrase, not a quotation); efforts “to nurture relations with the Environmental Protection Agency”; efforts “to challenge all unfair or erroneous treatment in the media”.
• Its credentials listed asbestos, vinyl chloride, fluorocarbons and dioxin, but not tobacco.
• The proposed lead, Matthew Swetonic, had worked for a cigarette manufacturer to convince the public that second-hand smoke was harmless and to “create a favorable public climate” for defeating smokers’ lung-cancer lawsuits. He had also done PR for Johns-Manville and been the first full-time executive secretary of the Asbestos Information Association.
• Once hired, it drafted letters for Brush’s ceramics customers “downplaying beryllium’s hazardous properties” (Marder, 1989b; Davis, 1989a,b). The Davis memos’ titles refer to a Defense Logistics Agency beryllium-oxide hazmat programme (p. 141).

6.3.4 Manufacturing uncertainty (pp. 136–138)#

• Private defection (p. 136). In 1989 Eisenbud, co-author of the limit and now a Brush consultant, told the company “he did not feel that he could defend the 2 microgram standard any longer”. This is second-hand: the source is a 1989 presentation by Brush vice-president R.H. Rozek (Rozek, 1989; reference p. 143).
• Liability shield (pp. 136–137).
• 1989: “Maintaining the existing [OSHA] standard is fundamental to successfully defending against any product liability litigation”.
• 1991 strategic plan (a board presentation; reference p. 143): “Employ legal means to defeat unreasonably restrictive occupational and emission standards”; “Resist an attempt to make the existing occupational exposure standard of 2 micrograms/cubic meter, as measured and calculated by Brush, more restrictive” (my emphasis: the plan defends the company’s own way of measuring compliance); and “The standard is safe, it is one of the most stringent standards, and it is fundamental to our product liability defense” (Rozek, 1991).
• DOE vs Brush (p. 137). The evidence “was clear to the DOE”, which began lowering the limit in 1991. Brush’s 1992 letter (Hanes, 1992a):
• DOE would “abandon the existing standard of over 40 years standing with no evidence” that it was unsafe or that the new one “affords any greater degree or [sic] safety”;
• NIOSH’s 0.5 had introduced “confusion”, which DOE’s “third number” would compound; “A proliferation of numbers as ‘standards’ can only weaken the acceptance, and therefore, the efficacy” of protection;
• “Confusion is never in the best interests of the worker”.
• The letter refers to DOE’s “new proposed standard”, so some DOE proposal existed by January 1992 (see Table 6.1 note).
• Internal DOE opposition (p. 137). Weapons offices argued “money spent protecting workers would mean less money for producing arms”. The rule sat “in limbo”.
• 1997 DOE forum (p. 137). Brush’s EHS director: “unaware of any scientific evidence that the standard is not protective”. Reports of disease below 2 μg/m³ were “sporadic” and “scientifically unsound” (DOE, 1997).
• The tautology (p. 137):
• Even without evidence of overexposure, overexposure was assumed “because the worker had developed CBD”, so the standard “must be fully protective”. The authors call this “the industry’s primary argument” and a “tautological construct”; “subsequent studies have demonstrated that the underlying logic to the argument was flawed”.
• 1991 talking points: (1) “Experience over several decades has, in our view, demonstrated” that exposures within the limit “afford a safe workplace”; (2) most employee cases involved exposure “(usually accidental), higher than the standard allows”, and where no overexposure could be found, “we also cannot say that there was not excessive exposure” (emphasis in original) (Pallum, 1991).
• The reference shows these notes came from a meeting on the “BLTT Program” (the blood test) and a draft titled “Efficacy of the 2 microgram per meter Standard” (p. 143).
• Collapse (pp. 137–138). DOE screening found “several hundred” CBD/BeS cases by the mid-1990s (DOE, 1998). “Scores” of workers sued Brush for failure to disclose toxicity.
• New line (p. 138). Continued denial “was unlikely to be a successful strategy”. Brush now “asserted that not enough was known to prevent CBD from occurring”. The authors add: “If true, the industry might avoid liability in CBD litigation” (their comment, not Brush’s words).
• By the time of the December 1998 DOE proposal, “Brush Wellman no longer asserted that the old standard was effective in preventing CBD” (p. 138).
• 1998 Brush–NIOSH collaboration (p. 138). “Productive, delivering findings that have substantially contributed to the literature” (NIOSH, 2002).
• DOE proposal and industry reply (p. 138):
• December 1998: DOE proposed an action level of 0.5 μg/m³ (“25 % of the OSHA standard”).
• February 1999: Brush’s Kolanz testified that “important research is underway which may provide a scientific basis for a revision”, citing particle size, number and surface area.
• Product defence (p. 138):
• Brush turned to Exponent, Inc., which the authors describe as providing “scientific and technical support to polluters and manufacturers of dangerous products” (citing its own 2006 annual report) and as “a leading practitioner of ‘product defence’”, defined as helping firms “reduce their regulatory burden and defeat liability claims” (Michaels, 2008).
• It convened a September 1999 conference co-sponsored by the American Conference of Governmental Industrial Hygienists (ACGIH), months before DOE’s final rule and just after OSHA signalled a revision.
• The proceedings (Paustenbach et al., 2001) called for research on particle size, compounds and skin exposure. The authors say they “went further, advocating postponement”, quoting: it “is difficult to identify a single new TLV”; “within three or four years, a series of TLVs might need to be considered”; the limit “could easily be among the most complex yet established”.
• Note: “advocating postponement” is the authors’ reading, not the paper’s wording.
• DOE final rule, December 1999 (p. 138). Protection “triggered at 0.2 μg/m³ rather than” the proposed 0.5, using the conventional full-shift mass measure.
• Statement of principle: “The government’s responsibility is to protect public health using the best available evidence. More research was, and is, needed”. Because the relationship of CBD to particle size, number and surface area “was, and remains, poorly understood”, the responsible officials decided policy “should not be delayed until this research was completed”.
• This is the authors’ voice describing a decision that Michaels himself directed (p. 132); it is advocacy by a participant as well as analysis.

6.3.5 New evidence but no new OSHA standard (pp. 138–139)#

• OSHA’s admission (Jeffress, 1998; written comments to DOE, 27 August 1998): “we now believe that our 2 μg/m³ PEL does not adequately protect beryllium-exposed workers from developing chronic beryllium disease, and there are adequate exposure and health effects data to support [the DOE’s] rulemaking”. The letter cited:
• CBD in machinists where “90 % of the personal exposure samples” were below the 0.01 μg/m³ detection limit;
• sensitisation risk “35–40 per 1 000 workers”;
• CBD risk among machinists “94 per 1 000”.
• Abandonment. OSHA committed to a new standard by September 2001 (OSHA, 2000). The incoming Bush administration “formally dropped its commitment … asserting it needed more information” (OSHA, 2002).
• Industry-linked evidence (p. 139).
• Studies by researchers “including several affiliated with Brush Wellman” show the standard does not fully prevent CBD (Henneberger 2001; Kelleher 2001; Schuler 2005; Stange 2001; Rosenman 2005; Madl 2007).
• The reference list identifies Kolanz as Brush’s Director of Environmental Health and Safety (p. 142), and he co-authored Madl et al. (2007). Other co-authors (e.g. Deubner) are not given affiliations in the section. Madl also co-authored the 2001 conference-proceedings paper (Paustenbach et al., 2001).
• A 2006 review “supported by Brush Wellman” conceded the OSHA standard “provides insufficient protection” (Borak, 2006).
• Downstream and community (p. 139).
• A 1999 sentinel case at a Quebec recycling plant led to 31 more cases at three plants (Robin, 2005).
• A follow-up survey found 2,789 Quebec workplaces using beryllium, including 63 golf club and 15 bicycle manufacturers (Tremblay, 2005).
• Eight new US community-acquired cases were identified in 1999–2002 (Maier et al., 2008).
• No safe level (p. 139). “There are many complex questions to answer, and there are relatively few workplaces in which these questions can be easily studied”. The National Research Council (NRC, 2008): “it is not possible to estimate a chronic inhalation-exposure level that is likely to prevent BeS and CBD”. NIOSH scientists: attempts to define a safe concentration “are not likely to be successful” (Kreiss et al., 2007).
• But act anyway (p. 139). This “does not, however, justify deferring” exposure reduction. Reducing exposure “to the lowest achievable levels” has cut BeS and CBD (NRC, 2008). ACGIH’s 2009 guideline value (TLV) is 0.05 μg/m³.
• Recommendation (p. 139). Substitute “whenever possible”. Where there is no substitute, “such as the production of nuclear weapons”, reduce exposure “to the lowest level technically feasible”.

6.4 Lessons for policymakers (pp. 139–140)#

Lessons are “widely applicable across many environmental health controversies”. Each is set in bold in the PDF, as is “manufacturing uncertainty” (verified from the PDF’s font runs).

1. “the absence of evidence is not evidence of absence”. The post-1950 lull in diagnoses “is likely attributable both to improved working conditions and the limitations of the diagnostic methods available at the time”. The BeLPT revealed hidden cases, “no doubt including cases that would not have been previously recognised as CBD”, but there were “indications before the advent of the BeLPT” that the standard “was not fully protective”. As low-exposure cases accumulated, “this conclusion became more difficult to avoid” (pp. 139–140).
2. “interpretation of scientific data by those with financial incentives for misinterpretation must be discounted”. - “It appears this incentive shaped the interpretation” by industry scientists, who defended the taxicab standard “long after it was correctly recognised as inadequate by independent scientists”. - Product-defence work “must be seen for what it is: advocacy, rather than science”. - “By the early 1990s, the accumulated evidence was sufficient”. Industry “manufactured and magnified uncertainty”. - “A more protective standard will help prevent CBD and save lives” (p. 140).
3. “considering the hazards associated with a toxic material throughout the entire life cycle of the product”. Hence “it would be prudent public health policy to end industrial use of beryllium, except in circumstances where substitution is impossible” (p. 140).

Table 6.1 Early warnings and actions (p. 140)#

As printed: - 1930s first uses and cases - 1949 AEC 2.0 - 1952 Registry - 1971 OSHA 2.0 - 1975 OSHA proposes 1.0, “never approved” - 1980s BeLPT - 1989 DOE proposes 0.5 for weapons and clean-up workers - 1999 DOE “issued a 0.2 µg/m³ exposure limit” - 2009 ACGIH 0.05 - 2012 “OSHA has yet to propose”

Discrepancies: - The text dates DOE’s start to 1991 and its formal proposal of a 0.5 action level to December 1998 (pp. 137–138); Brush’s January 1992 letter also refers to a DOE “proposed standard” (p. 137). “1989” matches neither date. It may be a transposition of 1998, but that is a guess. - The text calls 0.2 a protection “trigger”, not a limit (pp. 132, 138). - The table omits the 1951, 1966, 1972, 1974, 1977, 1994 and 1998 warnings documented in the text.

References (pp. 141–144)#

• Key evidence of intent comes from Bates-numbered or “available from authors” internal documents (Brush 1977; Hanes 1992a,b; Kohara 1974; Kolanz 1999; Pallum 1991; Powers and Preuss 1987; Rozek 1989, 1991; Tumbelson 1947). These are primary sources, but excerpted from a non-public, adversarial record.
• Minor errors:
• “Kriess” (p. 135);
• Hill & Knowlton cited “1986” but dated “17 November 1999” (p. 142);
• “Cronic” (caption, p. 134);
• Michaels and Monforton (2005) pages given as “S39–548” (p. 142).
• The cited 20/20 broadcast is titled “Thousands of Nuclear Weapons Works Exposed to Beryllium Dust are Sick or Dying” (ABC, 2000; p. 141). The chapter’s own text says “scores” of workers. The headline figure is not adopted or checked by the authors.

Panel: Guidotti (pp. 145–150)#

Opening (p. 145). - Inside the firm, corporate objectives align with “seemingly valid reasons to oppose change”, creating “a stronger argument from within than may be perceived from the outside”. - Hedged: “One suspects that most corporate leaders involved in situations like this live in a world of cognitive dissonance and denial rather than cupidity”. They believe their own denial, reinforced by “group-think, rationalisation, corporate culture, and a technical staff able and willing to provide justification”. - “The challenge is not to condemn the behaviour … It is to understand it in order to control it. (Prevention is probably not possible given human nature.)” - He explicitly responds to “Michaels and Monforton (2008 and present volume)”, i.e. both the journal article and this chapter (p. 145).

Case summary (p. 145). - Brush, “the only significant manufacturer and supplier of beryllium metal in the US”, first asserted adequacy, then “play[ed] on uncertainty”, challenging disease and then cancer data. - DOE “initially abetted” Brush, resisting “an evidence-based precautionary protective standard, presumably to protect the nuclear industry” (the section’s only “precautionary”). - OSHA had not acted “after at least 35 years of deliberation”. - Brush “was, and still is, a highly profitable company” with “a near monopoly”. “At the time it had little other business, although it is now more diversified” (relevant to how existential the stakes were). - He says the company “stands accused by Michaels and Monforton of cupidity and arrogance”. Those words do not appear in this chapter, nor (per a check of its online text) in the 2008 article; this reads as his characterisation, not a quotation. - His alternative reading: “pathway dependence”, in which an institution’s “initial worldview” commits it to a line “disgraceful for itself and tragic for the victims”.

Shareholder value (pp. 145–146). - He reads the chapter as assuming “throughout that the reluctance of Brush to accept new findings … was motivated by the desire to maintain corporate revenues” (p. 146). - US law and business culture prioritise shareholders (the 1919 Michigan ruling against Ford, i.e. Dodge v. Ford, whose significance Stout, 2008, argues “persuasively” in his view has been exaggerated; Friedman, 1970). The point is that “it was believed”. - Whether Brush broke the law is “outside the scope of this panel”. - “Fines are too low, sanctions are too weak, and legal actions too likely to settle”. Executives are “better off seeing the issue through to the end”, then “taking golden parachutes”. - Proposal. Recognise a fiduciary duty to “retreat from an untenable position”. Opportunities “to change direction without onerous consequences” would let leaders drop resistance “without admitting they were wrong”. He concedes it is “uncertain” whether this would change leaders’ course on their own initiative, though “likely” if consequences are high. What he calls “certain” is that legal recognition would give shareholders grounds to sue officers, and it fits “conservative business values … therefore hard to argue against” (p. 146).

National security and rationalisation (pp. 146–147). - “People tend to believe what is aligned with their own interest and do not recognise these beliefs as rationalisation”. - Early phase. A Cold War “risk/risk” calculus: catastrophic security stakes against harms that “seemed remote and uncertain”, judged “as they were understood at the time” (bold in original). “In hindsight, this was highly unfortunate but at the time it was not unreasonable”. - Later phase. “Risk/benefit” (the benefit of beryllium alloys). - Counterfactual. “In the context of an industry central to national security, it is not clear that the issue could have played out in any other way”, unless there had been a gradual transition to a more protective standard or earlier resolution of the uncertainties over the blood test (p. 147). - Stickiness. The 2 μg/m³ standard “was shown to be” inadequate “at the time”, but “took on a life of its own and became ‘sticky’” until the evidence became “overwhelming”. This “may have been, in part”, because the leap to 0.5 μg/m³ “was or appeared to be too great technically” and raised fears of disrupting the defence industry. - Main driver, in his view (hedged): reluctance “appears to have been driven mostly by uncertainties over the blood test (BeLPT), which was still in development and was perceived as unproven, a perception helped along by its alignment with financial interests”. - End phase. “Toward the end, it is clear that the emerging motivation for delay was perceived risk to the company”. He says “much of the opposition was self-serving”, but the company “could legitimately have been seen at the time” as a “critical industry”. That framing conflated “the company’s interests with the national interest” and gave a rationale he calls “(specious but persuasive)” (p. 147).

Acceptable risk (pp. 147–148). - The immune response is “stochastic”, so no standard is fully protective. The issue becomes “‘acceptable risk’ based on social criteria”. - Uncited norms: “one in a million” in society generally; “one in a thousand” in occupational health. - The company tried to decide “alone, without collective input … they did not [understand best], and they were working in a social vacuum”. Who decides: formally government, “informally … who has possession of the data”. - Defends the taxicab limit. He adds a detail not in the chapter: it was “revised upward at the last minute”. He concedes the story “suggests an overly casual approach”, but holds that without evidence for a threshold “a reasonable consensus based on informed opinion is not a bad substitute for data”, and “it certainly worked to eliminate acute Be disease” (like a Delphi group or a Bayesian prior). Judging it by later evidence is “like applying the legal and political standards of today to ancient history”. - Says DOE reduced its level “to 0.5 μg/m³ in 1998”. This was only proposed (p. 138). - Conflict of interest is “a given and a matter of degree” (p. 148).

Critique of the “discounting” lesson (p. 148). - He calls it a “blanket statement” and quotes it as “The interpretation of scientific data by those with financial incentives must be discounted”, without “for misinterpretation”. His wording matches the 2008 journal article verbatim (PMC2099330, checked 25 Sept 2026). So this is not a misquotation: this chapter’s version adds the narrowing qualifier, which partly meets his objection. - Incentives “attach to almost everyone”, including “those who receive support, to be critical of a position and offer expensive tests”. The real issues are “the degree of influence exerted on the investigator, the completeness of reporting, and the validity of the information”. Bias can affect academic studies too. Professional and personal incentives “may be as powerful”. - He credits “‘discounted’ rather than ‘ignored’”, but “in practice few who advocate tighter regulation may make the distinction”. Excluding corporate and grey literature “risks losing an immense body of valuable information”.

Transparency (p. 148). - Premise: “assuming that accountability is not possible in the adversarial setting of business interests, the issue reduces to one of transparency”. - “Can the data be audited and the analysis reconstructed? Who will ensure the integrity of the auditors?” - Health Effects Institute (HEI) reanalyses of air-pollution studies have “confirmed the original findings in every case”, and the process “is considered essential to acceptance of the findings by industry”. But the model is hard to apply to corporately sponsored research, which has “no external mechanism for guaranteeing quality assurance and no contractual obligation to cooperate with an audit”. - Trial registration and data repositories are further models, with “legitimate issues of business knowledge, anti-trust prohibitions, and proprietary information” and unclear enforcement. A voluntary approach by responsible firms would pressure the rest. - But these are “technical solutions and partial at best. The deeper issue is … commitment to a wrong decision in the face of diminishing room to manoeuvre”.

The blood test (pp. 148–149). - Accepting the BeLPT in 1983 would have shown due diligence, but consensus was lacking. - He cites “machinations of a cooperative academic”, attributed to the chapter authors but not in this chapter (nor, per a check of its online text, in the 2008 article). The source of the claim is unclear. - “Even today, the BeLPT has to be positive in two tests before a diagnosis is considered to be confirmed”. - “In the case of the BeLPT there was abundant evidence but the performance of the test was disputed”. Until validation, “it was not unreasonable for non-scientists to be sceptical” (my emphasis: the defence is limited to non-scientists). - “Who determines when a method is valid: scientists, regulators, or corporate interests? Likewise, who determines when data are sufficient and when evidence is actually absent?” - Delay was “logical (and, from their point of view, probably responsible)”. Over time the position “may have become untenable scientifically but the stakes were also rising”. “A late admission that the assay was valid would have been tantamount to admitting that the company had stalled and allowed further cases”. So “as uncertainty over the test diminished, the stakes increased, perhaps exponentially”, making it “paradoxically more attractive to defend their earlier position”. Leaders “probably truly did not believe” they had erred.

Lessons learned (pp. 149–150). - Disincentives: fear of liability and reputational damage, perceived fiduciary duty, shame, and “(probably inflated)” control costs. For beryllium these were buffer zones, containment and reassignment of sensitised workers. - Assuming leaders were honest “in their own terms (on the face of it, a problematical assumption)”, “there was never a moment at which Brush Wellman could change direction without paying what it considered to be an unacceptable penalty”. The company was “slogging through a tunnel of diminishing dimensions”. But “it would be going too far to characterise this as a ‘tragedy’ for them”. - Motives shifted: “At the beginning, the issue was cost (probably overestimated, as it usually is), loss of market share, and saving face. At the end, it may have been massive financial and legal liability” (p. 149). - The key question “may need to be recast as how information controlled by the company can be effectively accessed, with protection, so that the leadership has a viable escape route” (p. 149). - Precedent: Liggett & Myers’s 1996 break with the tobacco industry was “a business decision, pure and simple”. - Remedies: “there must be exit or escape opportunities”. These include forgiving legacy liabilities, a clear due-diligence defence, and “a threshold for sufficient knowledge”. The DOE compensation programme is “largely” an example. - The threat from dissident shareholders or hostile suitors, and institutional investors’ aversion to risk, adds pressure (pp. 149–150). - “If corporations are expected to reverse course, there must be room for them to turn around. Pressure builds resistance and ultimately denial and may be counterproductive at times” (qualifier matters). Escape routes for “bad corporations” are “not an attractive solution from the moral perspective but may lead to the greatest good for the greatest number” (p. 150). - Stickiness. A standard “develops a constituency and an infrastructure to support it”. “A policy of sequential standard-setting based on scientific evidence is inherently flawed. There will always be new scientific information.” Prefer “continuous improvement”. “Standards can be as much impediments as instruments of worker protection”. Continuous improvement is “perhaps unattainable in the current political context”, and its problems “lie in initiating it and sustaining it in an equitable manner” (p. 150). - Reference slips: “Pérez Carolli” in the text vs “Carrillo” in the list; Santo Tomas (2009) is listed but not cited.

Case timeline (if a case study)#

Lags (my reckoning from the chapter’s dates; the chapter does not compute them):

Harms. - “Scores” of weapons workers (pp. 131–132). - “Several hundred” CBD/BeS cases in DOE screening (p. 137). - Community cases (pp. 134, 139). - 31 cases from one Quebec sentinel case (p. 139). - No totals for cases, deaths or costs, and no counterfactual estimate. (The cited 20/20 broadcast’s title speaks of “Thousands” sick or dying, p. 141, but the chapter does not adopt that figure.)

What was known when.

The authors’ own lessons and conclusions#

Derived from their evidence (pp. 139–140):

1. Absence of evidence ≠ evidence of absence. Detection limits hid cases, and warnings pre-dated the BeLPT.
2. Discount interpretation by those with “financial incentives for misinterpretation”. Product-defence science is “advocacy, rather than science”, and the case illustrates “manufacturing uncertainty”. The underlying claim is hedged: “It appears this incentive shaped the interpretation” (p. 140).
3. Life-cycle hazards. Secondary users and recyclers are “unlikely” to manage exposure.

Stated as findings: - No safe level can be identified (p. 139). - Minimising exposure reduces BeS and CBD (p. 139). - The OSHA standard’s inadequacy is “confirmed” (p. 140). - The 1949 limit eliminated acute disease (p. 133).

Recommendations and advocacy: - Substitute “whenever possible”; otherwise reduce exposure to “the lowest level technically feasible” (p. 139). - “End industrial use of beryllium, except in circumstances where substitution is impossible” (p. 140). This is the strongest policy claim, made without analysis of substitutes, costs or benefits. - Use “the best available evidence” and do not delay pending research (p. 138). “This uncertainty does not, however, justify deferring implementation of programmes to reduce exposure” (p. 139). - A more protective standard “will help prevent CBD and save lives” (p. 140). This is unquantified.

Guidotti’s lessons (attribute separately): - Liability fear, reputational damage, shame, perceived fiduciary duty and “(probably inflated)” control costs make denial and rationalisation “the mode of behaviour” (p. 149). - Corporations need exit routes: forgiven legacy liability, due-diligence defences, a “threshold for sufficient knowledge” (p. 149). “Pressure builds resistance” (p. 150). - A fiduciary duty to retreat would harness shareholder pressure (p. 146). - Standards become sticky, so pursue continuous improvement (p. 150). - Transparency and auditability rather than exclusion of interested science (p. 148). - Acceptable risk is a social, not a corporate, decision (p. 147). “Who decides” validity and sufficiency is the deeper question (p. 148).

Mechanisms and dynamics#

1. Success against the visible harm masked the hidden one. - The limit eliminated acute disease and appeared to cut CBD, so it seemed “overly conservative” (pp. 133–134). - Chronic disease had a different biology: immune-mediated, idiosyncratic, possibly peak-driven. Its cases looked like anomalies and were explained away (p. 134). - Control of the dose-proportional harm was read as control of the hazard.

2. The measurement paradigm did not fit the mechanism. - Limits were full-shift mass averages (p. 138). Disease may track short peaks that standard methods “do not detect” (p. 134). - Machinists got CBD with most samples below detection (p. 139). - The mismatch later gave industry grounds to argue for waiting (particle size, number and surface area, p. 138); the authors concede these relationships were genuinely “poorly understood”. DOE acted with the imperfect metric anyway (p. 138).

3. Detection tools reshape the evidence base. - The BeLPT turned invisible sensitisation into countable cases: “several hundred” (pp. 135, 137). - Lesson 1 is largely about detection-limited false negatives, though the authors attribute the post-1950 lull partly to real improvements and stress that pre-BeLPT warnings existed (pp. 139–140). - Guidotti: new tools bring validity disputes, and who certifies them becomes contested (p. 148).

4. A provisional number hardened into an anchor with new functions. - “Tentatively” adopted (p. 133), then permanent, contractual and legal (1971). - Defended as “of over 40 years standing” (p. 137) and as “fundamental to our product liability defense” (p. 137). - Guidotti: “sticky”, with “a constituency and an infrastructure” (pp. 147, 150). - Revising it implied admitting past harm (my inference, supported by the liability memos, p. 137, and Guidotti’s “tantamount to admitting” argument about the blood test, p. 149).

5. Liability produced an escalation trap. - Internal memos tie the standard to litigation defence (pp. 136–137). The authors note that if the “not enough was known” line were true, “the industry might avoid liability” (p. 138), implying (without documenting) a strategic motive. - Guidotti: as uncertainty fell, stakes rose “perhaps exponentially”, making defence of the old position “paradoxically more attractive” (p. 149).

6. Denial, then doubt, then delay, with a fixed conclusion. - “Safe”, with cases dismissed as “unsound” (1991–97), then “not enough was known” and “important research is underway” (1998–99), then a Brush-supported review’s concession (2006) (pp. 137–139). - The premises shifted while the policy conclusion held constant. This pattern is central to the chapter’s argument (“a new rationale to delay”, pp. 131–132).

7. Engineering credibility. - A planned textbook whose memo stressed academic or not-for-profit auspices “to be fully acceptable and credible”, which appeared co-edited by the two Brush figures and an academic physician; a PR firm with asbestos and tobacco experience; a “definitive” white paper; outside scientists’ independent reviews (proposed); customer reassurance letters; a conference co-sponsored by a professional guideline-setting body (ACGIH); a product-defence consultancy (pp. 135–136, 138). - The common move is borrowing the authority of independent institutions. - Intent and outputs are documented. Effect on outcomes is not measured.

8. Private knowledge ran ahead of the public line. - NGK’s 1974 cases (p. 134) and Eisenbud’s 1989 disavowal (p. 136) vs “The standard is safe” (1991) and “unaware of any scientific evidence” (1997) (p. 137). - Litigation discovery exposed the gap (p. 132). Courts acted as a disclosure mechanism as well as a pressure.

9. The state was sponsor, customer, payer and regulator. - The AEC acted protectively but framed harm partly as a threat to its mission (p. 133). - Cost-plus contracting made the first limit affordable (p. 133). - Later, DOE weapons offices treated protection as competing with arms (p. 137), and DoD and DOE lobbied against OSHA (p. 135). Worker safety was the “last priority” (p. 132). - Guidotti: a Cold War risk/risk calculus, then a “critical industry” framing fusing company and national interest (p. 147). - Outcomes depended on which part of a divided state prevailed. DOE’s health office won in 1999 (pp. 137–138).

10. Political cycles gated action. - The 1975–77 proposal ended with the 1980 election (p. 135). OSHA’s 2001 commitment was dropped by the incoming administration (p. 139). - The regulator’s own written admission (1998) did not produce a rule (p. 139).

11. Burden and standard of proof. - Industry demanded proof that the old standard was unsafe and that a new one was safer (p. 137). It scrutinised each contrary case while assuming its own premise (overexposure) without proof (p. 137). - The authors: best available evidence; don’t wait for mechanism (pp. 138, 140). - Guidotti: there was no agreed “threshold for sufficient knowledge” (p. 149).

12. “No safe level” changes the regulatory logic. - Without a threshold, a limit expresses accepted residual risk. Guidotti calls this “acceptable risk” as a social choice (p. 147). - The chapter’s answer is minimisation and substitution (p. 139). Guidotti’s is continuous improvement (p. 150). - The original number may itself have rested on feasibility (Egilman, p. 133).

13. Hazard migrates down the value chain. - 1,500 of 28,000–107,000 exposed workers are in primary production (p. 132). 2,789 Quebec workplaces use beryllium, including golf club and bicycle makers (p. 139). - Recyclers, families and neighbours are exposed (pp. 134, 139). - Capacity is concentrated in the producer; exposure is diffuse. This is the basis of the life-cycle lesson (p. 140).

14. Who gains and who bears. - Gains: the weapons programme; a near-monopoly, “highly profitable” producer (p. 145) with cost-plus margins (p. 133); downstream product makers (p. 139). - Harms: production and bystander workers, downstream and recycling workers, families and neighbours (pp. 134–135, 139). - Control costs (buffer zones, containment, reassignment) are called “relatively expensive” but “probably overestimated” by Guidotti (p. 149). Neither author quantifies them. - Industry feared substitution most (p. 135). The authors recommend it (pp. 139–140).

15. Mental models and blind spots. - AEC founders: capable and confident, acting without data but provisionally, and the provisional status lapsed (p. 133). - Industrial hygiene: a threshold and dose-proportionality model made low-dose cases look like hidden overexposure (pp. 134, 137). - Company leadership: the standard as scientific truth and legal fortress (p. 137). Per Guidotti, cognitive dissonance, groupthink and belief in their own denial. They thought they “understood the problem best” while working “in a social vacuum” (pp. 145, 147, 149). - Weapons managers: a zero-sum budget view (p. 137). - Eisenbud’s trajectory: standard-setter, then industry consultant, then private critic of his own number (pp. 133, 136).

16. Framing and language. - “Public relations problem” (1947 AEC; 1989 H&K; chapter title) recasts harm as publicity. H&K’s “myths and misinformation” and “erroneous attitudes” recast the hazard as a false belief (pp. 133, 136). - “Safe”, “over 40 years standing” and “most stringent” use longevity and relative stringency as proxies for adequacy (p. 137). - “Confusion is never in the best interests of the worker” presents stronger protection as harmful to workers (p. 137). - “Sporadic” and “scientifically unsound” frame cases as isolated and defective (p. 137). - The authors’ counter-frames: “taxicab standard”, “product defence”, “manufacturing uncertainty”, “advocacy, rather than science” (pp. 133, 138, 140). - Guidotti’s: “sticky”, “tunnel of diminishing dimensions”, “room to turn around” (pp. 147, 149, 150).

17. Industry-linked science can also correct itself. - The Brush–NIOSH partnership (p. 138), Brush-affiliated studies and the Brush-supported 2006 concession (p. 139) show industry-linked science moving toward the truth in a collaborative setting. - This sits in tension with lesson 2.

18. Innovation and substitutes (thinly covered). - The BeLPT is the key innovation, a diagnostic one (p. 135). - Substitution is both the industry’s fear (p. 135) and the authors’ remedy (pp. 139–140). - Nothing on what beryllium enables, how substitutable it is, or the effects of regulation on innovation. Guidotti mentions only the “perceived benefit to society of Be alloys”, the “critical industry” framing, and fears that too great a technical leap would disrupt the defence industry (p. 147).

Transferable insights (technology-neutral)#

1. Controlling a hazard’s most visible, acute effect can create false confidence about a different, delayed effect with a different mechanism. - pp. 133–134, 139–140. - Strong: well documented, and the authors draw it explicitly.
2. Provisional limits set under ignorance can harden into anchors that acquire legal, contractual and reputational functions, making revision costlier than adoption. - pp. 133, 137, 147, 150. - Strong: internal documents tie the standard to liability defence.
3. Where harm depends on susceptibility and short peaks rather than average dose, standard metrics and thresholds can systematically miss it; “no safe level” shifts policy to minimisation and substitution. - pp. 132, 134, 138–139. - Strong for beryllium (NRC and NIOSH consensus). Moderate as a general claim, since the peak mechanism is explicitly speculative (p. 134).
4. Apparent absence of harm often reflects detection limits; new diagnostic or measurement tools can abruptly reveal hidden burden. - pp. 135, 137, 139–140. - Strong within the case. Guidotti adds that tool validity is itself contested (p. 148).
5. A stable conclusion defended by shifting premises (first that the standard is “safe”, then that not enough is known, then that research under way may justify a different limit) is a marker of motivated reasoning. - pp. 137–138. - Strong as documented behaviour. The motive (strategy vs sincere belief) is contested by Guidotti (pp. 146, 149).
6. Defences that attribute every observed failure to presumed breaches of a rule, never to the rule itself, are unfalsifiable. - p. 137 (talking points: “cannot say that there was not excessive exposure”). - Strong: directly documented.
7. Interested parties may borrow credibility by routing work through independent-seeming venues: academic co-editors or not-for-profit auspices, professional co-sponsors, outside reviewers. - pp. 135–136, 138. - Strong on intent (the 1987 memo) and on some outputs (the co-edited book, the co-sponsored conference). Moderate on the rest: the chapter does not show the book’s funding or whether the proposed independent reviews happened, and effect is unmeasured.
8. Calls for more research can be science or delay. The authors’ test is whether uncertainty is used to oppose protection that current evidence already justifies. - pp. 138–139. - Moderate: a useful criterion, but it requires judgment, and “advocating postponement” is the authors’ inference.
9. Liability exposure can make admitting risk costlier as evidence strengthens, producing escalating resistance. - pp. 136–138, 148–149. - Moderate: the liability link is documented; the escalation dynamic is Guidotti’s plausible reading.
10. When the state is sponsor, customer and regulator with an overriding mission, protection can be subordinated, and outcomes turn on internal divisions within government.
    • pp. 132–133, 135, 137, 147.
    • Strong: official admission plus documented internal opposition.
11. Arrangements that pass protection costs to a well-funded payer can ease early protective action.
    • p. 133 (cost-plus contracts “probably aided” acceptance).
    • Suggestive: asserted with “probably”.
12. Exposure migrates to diffuse downstream users and recyclers who lack producers’ capacity, so governance must cover the whole life cycle.
    • pp. 132, 139–140.
    • Moderate: the numbers are strong; the capacity claim is inferred (“unlikely”).
13. Regulatory action may follow political cycles more than evidence; a regulator’s own admission of inadequacy need not produce a rule.
    • pp. 135, 139, 140.
    • Strong on sequence. The weight of legal and procedural constraints is unexamined.
14. Private knowledge can run years ahead of public positions; compelled disclosure (litigation) may be the main route by which the gap closes.
    • pp. 132, 134, 136–137.
    • Strong for the documented instances.
15. Industry-funded science is not uniformly compromised. Collaboration with public agencies can yield evidence against the funder’s interest, and transparency or auditability may be a better safeguard than blanket discounting.
    • pp. 138–139, 148.
    • Moderate: supported by the chapter’s own evidence but in tension with its lesson 2.
16. Organisations committed to a position may need a face-saving, liability-limiting path to change course; otherwise rational leaders may persist in defending the untenable.
    • pp. 146, 149–150.
    • Suggestive: argument plus one analogy (Liggett & Myers); Guidotti concedes it is morally unattractive.
17. Standards treated as definitive snapshots invite stickiness; a pre-committed trajectory of continuous improvement may protect better.
    • p. 150; persistence of the 2 μg/m³ limit (pp. 133, 140).
    • Suggestive/asserted: untested, and Guidotti calls it perhaps politically unattainable.
18. Producers’ forecasts that a hazard classification will destroy them deserve scepticism.
    • The 1977 “no longer be a viable industry” claim was followed by Group 1 (1994) and “known carcinogen” (2002) classifications, and the firm is still “highly profitable” (pp. 135, 145).
    • Suggestive: a single case; survival is not proof of no economic effect; the feared stricter OSHA regulation had not arrived by 2012; and Guidotti notes the firm “is now more diversified” (p. 145).
19. Framing a health hazard as a “public relations problem” or as “misinformation” signals that perception, not exposure, is being managed.
    • pp. 133, 136.
    • Moderate: the language is documented, though the 1947 report also acknowledged “moral responsibility”.
20. Without agreed criteria for when a method is valid and when evidence suffices to act, delay has room to operate. Settling “who decides” in advance matters.
    • pp. 148–149; p. 140.
    • Moderate: well argued, but the validation dispute itself is thinly documented in the section.

Limitations, contestation and bias check#

Positionality and sourcing. - Michaels is historian and protagonist. He led the DOE rule the chapter praises (pp. 132, 138) and headed OSHA at publication (p. 131). - The chapter leans on his “personal knowledge” and on documents from litigation attorneys (p. 132), many “available from authors” (pp. 141–144). - These are real primary sources, but excerpted from an adversarial record, without surrounding context. - The company’s own account is absent (p. 145).

Advocacy vs analysis. - The case illustrates a thesis the authors developed elsewhere (pp. 131, 140), which risks fitting the evidence to the frame. - “End industrial use … except where substitution is impossible” (p. 140) is advocacy with no analysis of substitutes (availability, hazards, cost, performance) or of beryllium’s benefits. - Nuclear weapons are named as an example where there is “no adequate substitute” (p. 139). This is asserted without discussion, and substitutability is not assessed for any other use.

Causal attribution. - Delay is attributed mainly to the industry campaign plus defence and political priorities. This is well supported for DOE (pp. 132, 137). - The chapter does not examine OSHA’s own legal and procedural constraints on rulemaking or its resources. That is my observation; the section is silent. - The campaign’s actual effect on outcomes is asserted, not measured.

Genuine vs manufactured uncertainty. - The authors concede real unknowns: particle relationships “poorly understood” and no safe level estimable (pp. 138–139). - The line between caution and manufactured doubt therefore rests on use and motive. The key Paustenbach quote does not literally call for postponement (p. 138). - Guidotti: until validation, it was “not unreasonable for non-scientists to be sceptical” of the test (p. 148). He does not extend this to the company’s scientists.

The “discounting” lesson. - The chapter’s own evidence shows industry-linked researchers producing valid, adverse findings (pp. 138–139). - The authors’ qualifier “for misinterpretation” (p. 140) narrows the rule, but they give no way to identify such incentives in advance. - Guidotti’s critique is substantive (p. 148). He quotes the 2008 article’s wording, which lacked “for misinterpretation”. The EEA chapter’s qualifier narrows the lesson and partly answers him, but no text in the section acknowledges the change.

Hindsight. - The authors hedge the mechanism (“With hindsight, we can speculate”, p. 134) and praise the 1949 limit (p. 133). - Guidotti defends 1948–49 decisions as reasonable (pp. 147–148). - The strongest charges concern 1972–1999, when warnings were concrete, and are not weakened by the hindsight critique.

Inconsistencies and errors. - Table 6.1 “1989” vs the text’s 1991 start and 1998 proposal (pp. 137–138, 140). - 0.2 μg/m³ is called an “exposure limit” in Table 6.1 and a reduction by Guidotti (p. 147). The text more precisely calls it a protection “trigger” (pp. 132, 138), though it also refers loosely to the “current OSHA and DOE standards” (p. 139). Guidotti also says 0.5 was adopted in 1998; it was only proposed. - “Compelled OSHA to propose” (p. 135) conflicts with the 1975 date of the proposal; the cited 1977 NIOSH testimony was given at the hearing on that proposal (reference, p. 143). - Guidotti attributes words and claims not in this chapter: “cupidity and arrogance”; the “cooperative academic” (pp. 145, 148). He explicitly responds to both the 2008 article and “present volume” (p. 145), but a check of the 2008 article’s online text found neither phrase there either. “Cupidity and arrogance” is best read as his characterisation; the source of the “cooperative academic” claim is unclear. - Guidotti’s own tension: the limit “was shown to be” inadequate “at the time” (p. 147), versus judging it is “ancient history” (p. 148). This is probably 1949 vs later, but ambiguous. - Uncited assertions in the panel: the one-in-a-million / one-in-a-thousand risk norms; costs “probably overestimated, as it usually is” (pp. 147, 149).

Missing quantification. No totals of cases, deaths or compensation, no compliance costs, and no counterfactual estimate of harm avoidable by earlier action.

Case selection and scope. - Chosen as an exemplar of manufacturing uncertainty. - US-centred in an EEA report. The European operations and the UK study are only mentioned (p. 132), and there is no discussion of EU limits.

Currency. - References were accessed in 2011, so the text was apparently finalised around then. - The section does not mention the February 2012 joint draft standard from Materion and the United Steelworkers (see pointers below). By publication in 2013, the portrait of a still-resisting industry was partly out of date.

Fairness in both directions. - The chapter credits the AEC, the Brush–NIOSH partnership and industry-affiliated researchers (pp. 133, 138, 139). - Guidotti’s reframing adds insight but leans toward exculpation. He admits the honesty premise is “problematical” (p. 149), says “much of the opposition was self-serving” (p. 147), and declines to call it a “tragedy” for the company (p. 149). - He gives little evidence that the test dispute, rather than liability, drove resistance (“appears to have been driven mostly”, p. 147). His own account also names cost, market share, face-saving and, later, liability (p. 149). The chapter’s memos (Rozek 1989/1991) point to liability. - The “BLTT Program” origin of the 1991 talking points (reference, p. 143) suggests the two were intertwined.

Hindsight pointers (outside the section; primary sources checked 25 Sept 2026)#

• US OSHA final rule, 9 January 2017 (82 FR 2470):
• PEL 0.2 μg/m³ 8-hour TWA; STEL 2.0 μg/m³ over 15 minutes; action level 0.1 μg/m³ (29 CFR 1910.1024).
• Amended 2018 (83 FR 19948; 83 FR 39360) and 2020 (85 FR 42625).
• Estimated to “prevent 90 fatalities and 46 new cases of CBD annually once the full effects are realized”.
• The preamble records that in February 2012 OSHA received a joint draft standard from Materion Corporation and the United Steelworkers, in “substantial agreement” with OSHA’s proposal.
• It confirms the October 1975 proposal (40 FR 48814) rested on animal cancer studies and would have cut the TWA from 2 to 1 μg/m³.
• Sources: govinfo FR-2017-01-09, document 2016-30409; Cornell LII 29 CFR 1910.1024.
• OSHA exposure estimate: ~62,000 potentially exposed workers in ~7,300 establishments (osha.gov/beryllium).
• EU Directive (EU) 2019/983 (5 June 2019):
• binding limit 0.0002 mg/m³ (0.2 μg/m³, inhalable fraction);
• “dermal and respiratory sensitisation” notation;
• transitional value 0.0006 mg/m³ until 11 July 2026;
• the recital classifies beryllium as carcinogen category 1B (compare IARC Group 1, p. 135).
• Source: EUR-Lex, ELI dir/2019/983.
• Not verified; to check:
• the PEL proposed in the Materion–USW draft;
• whether Michaels was still OSHA head when the 2017 rule issued;
• the operation and cost of the DOE compensation programme (p. 149);
• post-2013 evidence on whether 0.2 μg/m³ prevents sensitisation.

Notable quotes#

1. “Priority one was production of our nuclear weapons … [the] last priority was the safety and health of the workers that build these weapons”. Richardson, 2000 (p. 132).
2. Public indignation “could reach proportions met with in the cases of silicosis or radium poisoning”. AEC, 1947 (p. 133).
3. “In the absence of an epidemiological basis for establishing a standard”. Eisenbud on the 1948 limit (p. 133).
4. “Seemingly trivial concentrations … no dose-response relationship appears to hold”. Stokinger, 1966 (p. 134).
5. If labelled a carcinogen, “it would no longer be a viable industry”. Brush Wellman, 1977 (p. 135).
6. “To be fully acceptable and credible … it will have to be published under the auspices of some not-for-profit organisation”. Powers and Preuss, 1987 (p. 136).
7. “The standard is safe … and it is fundamental to our product liability defense”. Brush plan, 1991 (p. 137).
8. “In these cases we also cannot say that there was not excessive exposure”. Brush talking points, 1991 (p. 137).
9. Product-defence science “must be seen for what it is: advocacy, rather than science”. Michaels and Monforton (p. 140).
10. “If corporations are expected to reverse course, there must be room for them to turn around”. Guidotti (p. 150).

Open questions#

1. What did US producers do with NGK’s 1974 report of below-limit cases (p. 134)? Was it disclosed to regulators?
2. How should the 1975–2017 delay be apportioned among the industry campaign, DOE/DoD priorities, changes of administration, and OSHA’s legal and procedural constraints?
3. What was the human cost of the delay, as a counterfactual using the later prevalence and sensitisation rates (p. 135)?
4. Why did the company move from resistance to concession (2006) and to a joint draft standard with the union (2012)? Does this support Guidotti’s exit-route thesis, or reflect other drivers (compensation schemes, litigation outcomes, the NIOSH partnership, management change)?
5. Did the DOE worker compensation programme function as the “way out” Guidotti describes (p. 149)? Who paid?
6. How and by whom was the BeLPT validated? What was the “cooperative academic” episode (p. 148)?
7. Does the peak-exposure hypothesis (p. 134) hold, and do short-term limits address it?
8. What did the UK investigation and Brush’s European operations show (p. 132)? Why did an EU binding limit come only in 2019?
9. Which uses are genuinely non-substitutable, and what are the substitutes’ own hazards and costs (pp. 139–140)?
10. How can interested science be “discounted” (p. 140) without losing valid data (p. 148)? Would audit or registration requirements have changed the trajectory?
11. Taxicab origin (Eisenbud) or feasibility (Egilman): which better explains the 1949 number (p. 133)?
12. Have downstream, recycling and community exposures fallen since 2013?

Audit log#

Audit of 25 Sept 2026 against the full extract (PDF pp. 133–152). Bold and italic runs were checked in the PDF’s font data. The 2008 Public Health Reports article was checked online (PMC2099330). Unless noted, changes apply to these notes; digest changes are listed at the end.

• Authors: “condenses the 2008 article” softened to “closely related to”; noted that the 2008 lesson 2 lacks “for misinterpretation”.
• Standpoint, summary box: added the box’s claim that the panel’s approach “may be better public policy”.
• Standpoint, panel dissent: “attributes resistance ‘mostly’” given its hedge (“appears to have been driven mostly”); added Guidotti’s “much of the opposition was self-serving”.
• Summary box: “secondary users and recyclers lack capacity” replaced with the verbatim “unlikely … have the expertise, resources and knowledge”.
• 6.1: Richardson’s “collaborated … to defeat” wording flagged as the authors’ paraphrase, not his words.
• 6.1: added that the DOE standard covers only weapons-complex workers.
• 6.2: added the 1947 report’s “more populous areas” point; labelled “lay knowledge as economic signal” as my gloss.
• 6.2: corrected Machle’s role (a physician consulting to the firm building Brookhaven, not an AEC scientist).
• 6.2: the threat to the AEC’s mission attributed, as in the source, to the accompanying “public relations problems”; the bold “costs passed to the state” claim softened to a labelled inference.
• 6.3.1: added the authors’ hedge on the registry evidence (“suggesting … might be occurring”).
• 6.3.1: added from the references that the 1974 NGK letter was addressed to a Brush vice-president.
• 6.3.1: “By 1989–91” corrected to the source’s “by the end of the decade”.
• 6.3.2: chronology note strengthened: the 1977 NIOSH testimony was given at the hearing on the 1975 proposal (references).
• 6.3.3: textbook: the chapter documents the plan and the co-editors, not the funding or not-for-profit auspices; publisher noted.
• 6.3.3: Hill & Knowlton’s “independent reviews” were wrongly quote-marked (the phrase is the authors’ paraphrase); added “echoed” 1947 and “common erroneous attitudes”.
• 6.3.3: Swetonic: added his role in the tobacco lawsuit-climate work, and “full-time”.
• 6.3.4: Eisenbud’s 1989 disavowal flagged as second-hand (reported in Rozek’s presentation).
• 6.3.4: 1991 plan: added the omitted “as measured and calculated by Brush” clause.
• 6.3.4: Hanes 1992 letter: added the “proliferation of numbers” argument and its reference to a DOE “proposed standard”.
• 6.3.4: tautology: added the authors’ “primary argument” and “logic … flawed” judgments and the first talking point.
• 6.3.4: “New line”: Brush’s assertion separated from the authors’ liability comment; added that Brush “no longer asserted” the old standard was effective.
• 6.3.4: Exponent: added the authors’ characterisation, sourced from Exponent’s own annual report.
• 6.3.4: DOE statement of principle: fuller reasoning added; flagged as the authors’ voice on a decision Michaels directed.
• 6.3.5: Jeffress letter: added date, recipient and the “adequate … data to support” clause.
• 6.3.5: Brush-linked co-authors: only Kolanz’s affiliation is given in the section; the claim about Deubner removed; Madl’s 2001 paper described accurately.
• 6.3.5: added the “relatively few workplaces” caveat to “no safe level”.
• 6.4: lesson 1’s hedge (“is likely attributable”) restored and fuller text added; bold formatting verified, including “manufacturing uncertainty”.
• Table 6.1: “probably a typo for 1998” softened to a guess; noted the 1992 “proposed standard”.
• References: added the “S39–548” slip and the 20/20 title’s “Thousands” (the body says “scores”).
• Panel opening: restored the “One suspects” hedge; added “(Prevention is probably not possible…)” and his explicit “(2008 and present volume)”.
• Panel case summary: added “only significant manufacturer”, “presumably to protect the nuclear industry” and “little other business”; “cupidity and arrogance” marked as his characterisation (not in the 2008 article either).
• Shareholder value: added his reading that the chapter assumes a revenue motive, Stout “persuasively”, legality out of scope, and his conceded uncertainty about the policy’s effect.
• National security: added his counterfactual, the “technical leap” explanation, the “appears” hedge, “self-serving” and “could legitimately have been seen”.
• Acceptable risk: added his concession of “an overly casual approach” and “worked to eliminate acute Be disease”; “revised upward” flagged as not in the chapter.
• Discounting critique: the “misquote” charge corrected (his wording matches the 2008 article); “paid” corrected to “who receive support”; added “few who advocate tighter regulation may make the distinction”.
• Transparency: “secured industry acceptance” corrected to “considered essential to acceptance”; added his caveats on corporate research and registries, and his premise.
• Blood test: restored “for non-scientists”; added “tantamount to admitting” and the two-positive-tests detail; “cooperative academic” also absent from the 2008 article.
• Lessons learned: added “not a ‘tragedy’”, the shift from cost and face to liability, and access to company-controlled information.
• Lessons learned: restored “and may be counterproductive at times”; fixed the “A policy of sequential standard-setting” quote; added “initiating it and sustaining it in an equitable manner”.
• Timeline: “sponsored textbook” corrected to co-edited; a 1998–99 row added for Brush’s shift; “Quebec (+31)” clarified as later findings.
• Lags: added Brush’s 1998–99 abandonment of the “effective” claim as an earlier end point (~24 and ~9–10 years).
• Harms: noted the 20/20 title’s “Thousands”, not adopted by the authors.
• Knowledge table: “industry insiders knew” softened to “were warned”.
• Recommendations: the non-verbatim “Uncertainty does not … justify deferring” replaced with the exact text.
• Guidotti’s lessons: disincentives list completed, with “the mode of behaviour” quoted.
• Mechanism 2: “legitimate-sounding foothold” replaced with neutral wording plus the authors’ concession that the unknowns were real.
• Mechanism 3: “Lesson 1 is really about” softened.
• Mechanism 4: “admitting past harm” labelled as my inference, with supporting pages.
• Mechanism 5: “adopted because it might avoid liability” softened to the authors’ “if true” remark.
• Mechanism 6: non-verbatim quotes fixed; “core evidence” softened.
• Mechanism 7: textbook description corrected; ACGIH identified.
• Mechanism 18: added Guidotti’s “critical industry” and defence-disruption points.
• Insight 5: non-verbatim quoted phrases replaced by paraphrase.
• Insight 7: “academic publishers” corrected; rating split between documented intent and unshown outputs.
• Insight 18: added caveats (no OSHA rule by 2012; diversification).
• Limitations: nuclear-weapons exemption reworded (the source gives “no adequate substitute” as the reason).
• Limitations: Guidotti’s “not unreasonable” scepticism limited to non-scientists.
• Limitations: discounting-lesson “misquote” corrected.
• Limitations: inconsistencies updated (the text’s loose “standards”, the 1977 testimony, Guidotti’s attributions checked against the 2008 article).
• Limitations, fairness: added Guidotti’s concessions and his own list of motives.
• Digest: Eisenbud and Machle described correctly; Egilman’s feasibility reading added; “eliminated” corrected to “virtually disappeared”.
• Digest: “family members” replaced by the source’s non-worker examples; NGK letter to a Brush VP added; BeLPT dates corrected.
• Digest: OSHA 1975 “killed” softened to “never finalised”; textbook corrected; “not enough is known” de-quoted; Eisenbud’s disavowal flagged as second-hand.
• Digest: lesson 1 nuance and lesson 2 hedge added; Quebec sentence corrected (the survey, not the case, found 2,789 workplaces).
• Digest: Guidotti hedges and concessions restored; the 2008 wording of lesson 2 explained.
• Digest: “forecasts of ruin proved wrong” softened; inconsistencies rewritten; outside-source items labelled.
• Checked and confirmed without change: all figures (28,000–107,000 and 1,500; 20 and 36 cases; 0.1–4.4 %, 0.9–9.9 %, 9–100 %, 6–8 %/yr; 90 % below 0.01; 35–40 and 94 per 1,000; 31 cases, 2,789, 63 and 15; eight community cases; 0.05 TLV); all dates in 6.3.x; the other quotations; Table 6.1 as printed; the reference slips listed earlier.