# LATERAL RESEARCH — forgotten components of the story (2026-06-11)

*Goal: find missing important components that, if added, enrich the "numerology of toxicology" essay.
Method: 5 parallel lateral-research agents, each chasing one adjacent round-number / threshold thread,
each given the current 12-section kit so they hunt gaps, not duplicates. Grades per CLAUDE.md:
**[CONFIRMED-primary] / [secondary] / [absence-of-evidence]**. Locators are only those verified; where a
cite still needs a primary pull before going in-text, it says so.*

---

## STATUS UPDATE 2026-06-11 — papers collected
**9 open-access/government primaries fetched, title- AND content-verified, in `papers/lateral/`** (catalog:
`papers/lateral/_LIBRARY.md`; paywalled remainder: `papers/lateral/_WISHLIST.md`; excerpts under `sources/`).
Grade flips this buys:
- **10⁻⁶ origin → now [CONFIRMED-primary] (held):** Rodricks 2019 (firsthand: Hutt's "1 in 1 million"; Mantel's
  "virtually safe" 10⁻⁸) + Crump 2018 (single-cell/no-threshold; Mantel-Bryan procedure). The linchpin (0.5 ppb
  = sub-10⁻⁶) was already primary-confirmed in the local 1995 FR.
- **Ames → [CONFIRMED-primary] (held):** the 1973 Ames test + both 1990 PNAS papers (99.99%; "insignificant";
  "half are rodent carcinogens").
- **RfD tower → [CONFIRMED-primary] (held):** EPA 2002 RfD/RfC process review (the ×10 default UFs).
- **GRAS → [CONFIRMED-primary] (held):** GAO-10-246 + 81 FR 54960 + NRDC 2014 (the structure + "without FDA's
  approval or knowledge" + voluntary notification). *Caveat:* the JAMA 2013 COI percentages stay wishlisted.
- **Still wishlisted (paywalled / pre-1994 FR / NAP-gated):** Mantel & Bryan 1961; the FDA SOM Federal Register
  pages; Rulis 1992; Kelly 1991; Travis 1987; Ames 1987 *Science* (HERP table); Perera & Boffetta 1988; NRC
  1996; Dourson 1983/88; Lehman & Fitzhugh 1954; Castleman & Ziem 1988 + Roach & Rappaport 1990 (TLV — none held
  yet); Neltner 2013. All within reach of operator `doc_retrieve`/library; DOIs in `_WISHLIST.md`.

---

## HEADLINE — the one component the essay is built to need and doesn't have

### The "one in a million" (10⁻⁶) acceptable cancer risk — the carcinogen-side de minimis

**Why it is THE missing keystone, not just another addition.** §9 ("The numbers that didn't hold") makes the
essay's sharpest point: for a genotoxic carcinogen there is no dose that does nothing — "the number names a
quantity the chemical doesn't have." The essay stops there, at the Delaney wall. But regulators did **not** stop
there. Where a no-effect level cannot exist, they built one anyway, out of a different round number: a lifetime
cancer risk of **one in a million**. It is Frawley's exact move — a round number standing in for a threshold the
science denies — performed on the one class of chemicals the essay says has no threshold. The story currently
sets up the void and never shows it being filled.

**The linchpin fact — verified today against the operator's own local primary** (`sources/FDA_Threshold-of-
Regulation_60FR36582_1995.txt`, lines 277–386). The essay already discusses the 1995 Threshold of Regulation
rule and its 0.5 ppb, but describes the number only as "two hundred times under Frawley's figure." It never says
**why 0.5 ppb**. The rule says why, in its own voice: **[CONFIRMED-primary]**

> "FDA has determined that, if an exempted substance present in the diet at 0.5 ppb were later found to be a
> carcinogen, the upper-bound lifetime risk resulting from the use of the substance is likely to be **below one
> in a million.**" (60 FR 36582 at 36583, 17 Jul 1995.)

> "The agency plotted the potencies [of 477 animal carcinogens] as a probability distribution on a
> semilogarithmic scale and found that they formed a **bell-shaped distribution curve.** Using this probability
> distribution … FDA determined that most known carcinogens pose less than one in a million upper-bound lifetime
> risk if present in the daily diet at 0.5 ppb or less." (same, 36583.) **[CONFIRMED-primary]**

So the de minimis number for "trifling" food-contact chemicals — the rule the essay already builds its §11
around — is **literally a 10⁻⁶ calculation read off a bell curve.** And a footnote hands the numerology theme a
gift: the rule notes its 0.5 ppb "happens to be 200, rather than a 100, times lower" than the toxic level, "is
**merely coincidental** and does not reflect the agency's reasoning" — while footnote 1 simultaneously invokes
"a 100-fold safety factor … (1000-fold to 2000-fold)." The first essay's factor of 100 is sitting right there in
the footnotes of the rule that closes this one. **[CONFIRMED-primary]**

**Where 10⁻⁶ came from (the origin story — the prize for the essay).**
- **Mantel & Bryan (1961),** "Safety Testing of Carcinogenic Agents," *J. Natl. Cancer Inst.* — NCI
  biostatisticians defined a "virtually safe dose" as one giving a lifetime risk of **1 in 100 million (10⁻⁸)**,
  via log-probit extrapolation; prompted by the 1959 cranberry/aminotriazole scare. Asked how he picked it,
  Mantel reportedly said, **"We just pulled it out of a hat."** — 10⁻⁸-in-1961 is **[CONFIRMED-primary]**
  (corroborated by Crump 2018, *Dose-Response*); the "hat" quote is **[secondary]**, reported anecdotally by
  Kelly (below) with no footnote — usable, but attribute to Kelly, do not grade primary.
- **The DES / "Sensitivity of the Method" seam (the carcinogen statute grows its OWN de minimis from the
  inside).** The **1962 DES proviso** amended the Delaney Clause to allow a carcinogenic animal drug if **"no
  residue"** is found by approved methods (21 U.S.C. §348(c)(3)(A)). When detection got too good in the 1970s,
  FDA (General Counsel **Peter Barton Hutt**) redefined "no residue" not as zero but as a **calculated** level: a
  residue below the dose giving a **10⁻⁶ lifetime cancer risk** — "a quantitative criterion of insignificance."
  Proposed 1973 (originally 10⁻⁸), settled at 10⁻⁶ by 1979, final SOM rule **52 FR 49572 (31 Dec 1987).** This
  is quantitative cancer-risk assessment's birth in U.S. law — and it is Frawley's trick performed **through**
  the absolute statute, a decade before the courts (*Public Citizen* 1987, *Les* 1992) said Delaney admits no
  trivial-risk exception. — DES proviso + §348(c)(3)(A): **[CONFIRMED-primary]** (DOJ/OLC). The 10⁻⁶-criterion
  and Hutt's authorship: **[secondary-authoritative]** (NRC *Risk Assessment in the Federal Government*; Rodricks
  2019, PMC6366000). FR page spans (38 FR 19226; 42 FR 10412; 44 FR 17070) carried from NAS/Kelly — **pull one
  directly before quoting a page.**
- **Kelly, "The Myth of 10⁻⁶ as a Definition of Acceptable Risk"** (Air & Waste Mgmt. Assn., 1991) — phone-
  surveyed the federal "Who's Who" who use the number; none could document its origin. Thesis (read firsthand):
  the number "has no scientific or regulatory basis. Its use appears to be arbitrary." Officials' verbatim
  confessions: *"I just assumed it was because one-in-a-million sounded like such a nice phrase"* / *"It was
  selected because it was 'doable'"* / *"based on the chance of being hit by lightning"* / *"a purely political
  decision made … behind closed doors in the 1970s"* / *"You really shouldn't be asking these questions."*
  **[secondary]** — the authoritative investigation of the question; quotes read firsthand from the PDF.
- **EPA hardened it into the 10⁻⁶–10⁻⁴ "acceptable range"** with 10⁻⁶ as the "point of departure," Superfund
  NCP final rule, **40 CFR 300.430 (1990).** **[CONFIRMED-primary]** for the range/text; Travis et al. 1987
  found the *de facto* acceptable level across 132 federal decisions tracked nearer 10⁻⁴ — the number moved with
  the size of the exposed population and politics, not toxicology **[secondary]**.

**Slots into:** deepen **§11 ("It became law")** with the one-in-a-million derivation of 0.5 ppb (one or two
sentences + the footnote-100 gift), **and** a short new block — working title **"The number they pulled out of a
hat"** — between §9 (no-threshold void) and §11 (legal lineage), carrying Mantel, the DES/SOM seam, and the
officials' confessions. Pitch: *Where a no-effect level can't exist, regulators built one anyway — 10⁻⁶ — and
its own authors say they pulled it out of a hat. It is the carcinogen-side twin of Frawley's 0.1 ppm.*

---

## RUNNERS-UP (ranked) — genuinely missing, each enriches; operator's call on how many to use

### 2. Bruce Ames — de minimis scaled to the whole diet (strong late section / coda)
The man who built the **Ames test** (Ames et al., *PNAS* 70:2281, 1973) — the cheap mutagen→carcinogen screen
that operationalized the **no-threshold** paradigm, the very science that should have killed "toxicologically
insignificant levels" — became, a decade later, the apostle of "it's all trivial." Ames, Magaw & Gold, **"Ranking
Possible Carcinogenic Hazards," *Science* 236:271 (1987)** introduced **HERP** (Human Exposure / Rodent Potency):
a peanut-butter sandwich or a glass of wine outranks the synthetic residues; "carcinogenic hazards from current
levels of pesticide residues or water pollution are likely to be of minimal concern relative to … natural
substances." Ames, Profet & Gold, **"Dietary pesticides (99.99% all natural)," *PNAS* 87:7777 (1990)**: "99.99%
(by weight) of the pesticides in the American diet are chemicals that plants produce to defend themselves … the
comparative hazards of synthetic pesticide residues are insignificant." Frawley said *this additive* is a trifle;
Ames said the *entire diet* is — de minimis at cosmic scale, in a new round number (99.99%). And the NAS lends its
name **again**: NRC, *Carcinogens and Anticarcinogens in the Human Diet* (1996) adopts Ames's "same yardstick"
framing while declining his policy punchline. **All four papers verified [CONFIRMED-primary]** (DOIs/PMIDs
checked). The connective tissue to our story: Frawley was an author in the **inaugural June 1981 issue of
*Regulatory Toxicology and Pharmacology*** — the ISRTP "sound science" venue where Ames's argument lived
(alongside Coulston, Kolbye, Lederberg) **[secondary — confirm the Vol.1 No.1 TOC for a primary flip].** No
direct Frawley↔Ames citation found — frame as parallel apostles of one logic in one milieu, **do not overstate**.
Critiques to carry for fairness: HERP conflates exposure-ranking with risk (Perera & Boffetta 1988); rebuttals in
*Science* by Stoloff (FDA's aflatoxin authority) and Neutra ("outrage factor"). **Slots:** new coda-adjacent
section after §9, before/around §11/§12.

### 3. The RfD / uncertainty-factor tower — the factor of 100, metastasized (bridge to essay one)
EPA's **Reference Dose** = a NOAEL ÷ a *product* of round tens: ×10 interspecies × ×10 intraspecies (=100), ×10
subchronic→chronic, ×10 LOAEL→NOAEL, ×10 database → divisors of 100; 1,000; 10,000. **Barnes & Dourson (1988),**
*Regul. Toxicol. Pharmacol.* 8:471, building on **Dourson & Stara (1983),** 3:224 (PMID 6356243), trace the
practice **explicitly back to Lehman & Fitzhugh (1954)** — the exact three-page editorial the *first* essay is
built on — and concede the original factor selection "appears to have been rather arbitrary." EPA's own IRIS
definition calls the RfD "an estimate (with **uncertainty spanning perhaps an order of magnitude**)." The factor
of 100 didn't just survive; it became an official tower of stacked tens running U.S. environmental regulation.
**[CONFIRMED-primary]** on the citation chain + EPA's "order of magnitude" admission; **verify the "rather
arbitrary" wording against the 1983 PDF before quoting verbatim.** **Slots:** a bridge beat in §11, or near §8.

### 4. GRAS self-certification — naming the mechanism Frawley was actually working (sharpens the spine, not a new fact)
The dossier already has the quote; it never names the machine. At the Feb 1968 conference Frawley said lawyers
told him that two dozen experts writing in "confirms that these uses are generally recognized as safe or 'gras'
and that no action on the part of the FDA is necessary" (`papers/fdclj_1968.txt`) **[CONFIRMED-primary].** GRAS
(FFDCA §201(s), created by the **same** 1958 amendment he was trying to escape) lets a documented consensus of
"experts qualified by scientific training and experience" exempt a use from FDA pre-approval (21 CFR 170.30). So
Frawley's "24 toxicologists in writing" was not a plea for a rule — it was a **GRAS self-certification**:
manufacture the expert consensus, declare it safe, skip the test. The modern endpoint makes the COI rhyme exact:
**Neltner et al., *JAMA Intern. Med.* 2013** — of 451 GRAS determinations, **22.4% made by an employee of the
manufacturer, 64.3% by a manufacturer-selected panel, 0% by an independent party** **[CONFIRMED-primary]**;
**GAO-10-246 (2010)**: companies "may determine a substance is GRAS without FDA's approval or knowledge"
**[CONFIRMED-primary]**; notification stayed voluntary through the 2016 final rule (81 FR 54960). Caveat: in 1968
the voluntary-notification regime didn't exist yet — Frawley invoked the GRAS *concept*; say "concept" to avoid
anachronism. **Slots:** sharpen §3/§4 (the conference) and §6 (the COI pattern) — fold in, don't add a 13th
section.

### 5. TLVs / Castleman–Ziem — the occupational mirror of de minimis (narrative rhyme with vinyl chloride)
ACGIH's **Threshold Limit Values** (first published 1950) are the occupational round numbers. **Castleman & Ziem,
"Corporate Influence on Threshold Limit Values," *Am. J. Ind. Med.* 13:531 (1988)** found unpublished corporate
communications shaped TLVs for **104 substances — 15 based on that alone** — with Dow/DuPont "consultants" writing
the TLVs for their own employers' chemicals. Same move as de minimis: industry sets the "safe" number, then it
becomes the standard. Connects directly to §9's vinyl chloride (which had a comfortable TLV before it had a cancer
finding). **[CONFIRMED-primary]** on the citation + findings; pull the verbatim sentence before quoting.
**Slots:** a short beat beside §9.

### Minor / connective (use a line at most)
- **1959 Cranberry scare** (Flemming, aminotriazole, first Delaney-era panic) — vivid cold-open color, BUT the
  **Hercules link is NEGATIVE** (maker was American Cyanamid; Hercules's herbicide line is 2,4,5-T/2,4-D). Use as
  scene-setter only. **[secondary; well-sourced negative on Hercules].**
- **Precautionary principle / Wingspread 1998** — the explicit COUNTER-POLE: "the proponent … should bear the
  burden of proof," the exact inversion of presumed insignificance. Names the opposite of the Delaney/Dingell
  frame. **[CONFIRMED-primary]** (SEHN text).
- **1983 NRC "Red Book"** — the four-step risk paradigm that turned round-number guessing into "method." One-
  paragraph hinge, not a section.
- **Hormesis / Calabrese** — low-doses-are-good, industry-funded; one line in §10/coda at most.

---

## What to pull to primary before any of this goes IN-TEXT
1. The **Mantel "pulled it out of a hat"** line — attribute to Kelly; it is anecdotal, not primary.
2. One **Federal Register** page of the DES/SOM rulemaking (38 FR 19226 / 42 FR 10412 / 52 FR 49572) to quote the
   "no residue"/10⁻⁶ criterion verbatim. (The 1995 ToR linchpin is already primary-confirmed locally.)
3. **RT&P Vol. 1 No. 1 (June 1981) table of contents** to flip the Frawley↔Ames-milieu link to primary.
4. **Dourson & Stara (1983)** PDF for the "rather arbitrary" wording; **Castleman & Ziem (1988)** for its verbatim
   "solely … unpublished corporate communications" line.

*Agent transcripts: 5 parallel general-purpose agents, 2026-06-11. Linchpin (0.5 ppb = sub-10⁻⁶ calc) re-verified
by me against `sources/FDA_Threshold-of-Regulation_60FR36582_1995.txt` ll. 277–386.*