# CLAUDE.md — John P. Frawley project

*Project charter and working instructions. Detail lives in the numbered dossier files (`01`–`10`); this file
is the objective, the scope, and the conventions. Last updated 2026-06-11.*

## What this project is

A documentary-history investigation of the toxicologist **John P. Frawley** (FDA Division of Pharmacology →
chief toxicologist at Hercules Inc.), built toward a **"Reckoning Science" essay** — the sequel to the first
piece in the series, *"Of Mice and Men and Round Numbers,"* which traced the 100-fold safety factor. The
unifying theme is the **numerology of toxicology**: round numbers (the factor of 100; Frawley's 0.1 ppm) used
as stand-ins for knowledge we don't have.

The deliverable is the essay, drafted as **movable sections** in `reckoningscience/sections.md` (rendered to
`sections.pdf`). The earlier continuous draft `reckoningscience/fda-er_said_sheer_nonsense.md` is **superseded**;
keep it only for reference. The dossier files exist to *support* the essay with verified primary evidence.

A connected **side-analysis lives in `ssd/`**: it carries the same theme into **ecological** regulation — the
Species Sensitivity Distribution and its 5th-percentile **HC5** — the *same* round-number device (5th percentile
+ a safety factor) applied to ecosystems, and ties that lineage back to de minimis. It is its own body of work
(own library and analysis docs), feeding the series rather than the single Frawley essay.

## Objectives (what we are actually trying to establish)

1. **The "ignorance factor."** Frawley's own 1965 description of the 100-fold margin as an "ignorance factor…
   overly conservative" — the bridge from the first essay and the seed of everything after.
2. **The de minimis / "toxicologically insignificant levels" proposal:** its commercial driver (the rosin
   testing bill), its content (0.1 ppm in diet / 0.2% in container), and how it was derived (the ignorance
   factor, plus "another factor of ten").
3. **The FDA reaction:** the Feb 1968 National Conference; Summerson's "sheer nonsense"; the organized
   industry campaign behind the "24 toxicologists."
4. **Institutional endorsement:** the 1969 NAS *Guidelines for Estimating Toxicologically Insignificant Levels*
   — Frawley's phrase, his number, his rationale, drafted by a task force that **included him**.
5. **The IBT relationship.** How **close** Frawley was to Industrial Bio-Test Labs (named Calandra
   co-authorships; Hercules contracting), and — separately — whether his de minimis dataset rests on IBT's
   fraudulent work (analysis in `09`: largely refuted; the genuine vein is the Hercules↔IBT closeness, not the
   dataset).
6. **The conflict-of-interest pattern:** co-author of his contract lab; employer chaired the lobby; seated on
   the NAS panel that blessed his proposal — never disclosed beyond the by-line.
7. **The dioxin / Agent Orange "other file":** the two registers (public "insignificance" vs the private
   Farnham memo); the 1983 summary judgment won on his testimony, later vacated.
8. **Hindsight failure of the no-effect levels:** vinyl chloride, acrylamide, cyclamate — the genotoxic
   no-threshold "category error."
9. **The 1981 manifesto:** his answer to the IBT fraud was to attack the *regulators* (GLPs, Delaney, "zero
   risk").
10. **Propagation into law:** Frawley → Rulis (1987) → the 1995 Threshold of Regulation (21 CFR 170.39) →
    TTC / EFSA (2019); and the parallel **judicial de minimis doctrine** — *Monsanto v. Kennedy* (1979,
    Leventhal) → *Public Citizen v. Young* (1987, Williams) → *Les v. Reilly* (1992), where de minimis hits the
    Delaney wall — in `10_DE_MINIMIS_LEGAL_LINEAGE.md`.
11. **The ecological parallel (side-analysis in `ssd/`):** the SSD / HC5 as the same 5th-percentile threshold
    device for ecosystems — its **genealogy** (US EPA Stephan/Mount + Dutch RIVM/VU/Danish NERI: government-and-
    academic-born, *not* industry-pushed, unlike de minimis); its **thin, circular validation** (Versteeg 1999
    = Procter & Gamble; Hose 2004; ECETOC mesocosm — Mount & Stephan's own 1967 "verify by field observation"
    caveat never closed); and the **two-pole framing** for the series — **Delaney** (the absolute / no-safe-dose
    statute) vs. the **calculated threshold** (Frawley's de minimis, the HC5, Dingell's procedure-and-balance
    governance). The round number won, even where the no-threshold science was right.

## Scope of biography

**Professional biography is in scope.** Pursue his career and professional formation: FDA tenure (Division of
Pharmacology), education and degrees (as credentials), Hercules role and title progression, professional bodies
(SOT charter member, ISRTP, his consultancy), and his **military service** — the contested 1953 "46th MASH /
1st Lt. J. P. Frawley, Medical Service Corps" question. Military service is **in scope** because he may have
served in a *professional* (medical / shock-physiology / research) capacity, which bears on his career arc and
the 1953–55 publication gap; resolving that identity is a legitimate objective.

**Out of scope — personal matters only.** Do not spend effort on: death date / obituary / "in memoriam"; birth
date/place; family. These add nothing to the professional story.

In `05_OPEN_QUESTIONS.md §B`: **B1 (death/obituary) is out of scope**; **B2 (1953 MASH military service) and
B3 (education / title progression) are in scope.**

## Conventions (follow these)

- **Evidence grading on claims:** `[CONFIRMED-primary]` / `[secondary]` / `[absence-of-evidence]` (register
  uses `[P]`/`[P-cite]`/`[2]`/`[abs]`). **Never invent a locator** (Bates, DocumentCloud ID, NAL item, DOI,
  call number). If you can't confirm a specific id, name the archive generically.
- **When the operator drops a new primary PDF into `papers/`:** read it in full (`pdftotext -layout`); write a
  short quotable excerpt to `sources/` + a line in `sources/_SOURCES_INDEX.md`; then integrate across the
  numbered files (register grade, bibliography, timeline, master dossier, analytic memo, open questions). Flip
  `[secondary]→[CONFIRMED-primary]` wherever the new primary backs a prior second-hand claim.
- **Quotes:** in the dossier keep verbatim quotes short; in the essay they may run longer but must be exact
  (clean only obvious OCR). `sections.md` is currently a **quote-rich working version** — quotes are
  intentionally over-long and get trimmed at assembly.

## Reckoning Science voice (for the essay)

- Tight, literary-investigative prose; **relatively short sentences; avoid AI tics** (no portentous balanced
  one-liner at every section end; no over-balanced "not X but Y"; no over-signposting).
- **Structure serves *this* story** — do not mechanically mirror the first essay's architecture. Write good
  self-contained story blocks; sequence them afterward (the operator decides order). Each fact/quote lives in
  exactly one block — **no repetition** (this is the standing #1 complaint).

## Repository layout

- `01`–`10_*.md` — the dossier (master, bibliography, document register, timeline, open questions, analytic
  memo, ToxicDocs findings, vinyl-chloride campaign, IBT↔Frawley overlap, **judicial de minimis legal lineage**).
- `papers/` — primary PDFs + their OCR `.txt`. `sources/` — downloaded sources, `.md` excerpts, the retrieval
  list. `toxicdocs/` — 251 downloaded "Frawley" docs.
- `reckoningscience/` — the essay (`sections.md`/`.pdf`), plus full-draft variants (`de_minimis_draft*.md`) and
  candidate openings (`framing_poles_delaney_dingell.md` — the Delaney/Dingell "absolute vs. round number" frame).
- `ssd/` — the **SSD ↔ de minimis side-analysis**: a 20-paper **title-verified** library (`papers/_LIBRARY.md`,
  `_WISHLIST.md`), regulatory primaries (`sources/`: EPA 1985 guidelines, ECETOC WR28, the 1965 Water Quality
  Act), and analysis docs each rendered to `.pdf` — `SSD_GENEALOGY`, `SSD_ORIGINS_AND_AFFILIATIONS`,
  `SSD_VALIDATION`, `SSD_MANDATE_AND_POLITICS`, `SSD_DE-MINIMIS_LINEAGE`.
- `frawley.bib` / `frawley.csl.json` — machine-readable bibliography.

## Tooling notes

- **Obtaining papers:** **not via Sci-Hub** (copyright infringement). Lawful routes (institutional access via
  DOI, ILL, HathiTrust/IA for the public-domain pre-1964 FDA papers) are listed in
  `sources/_FRAWLEY_PAPERS_RETRIEVAL.md`.
- **MD → PDF:** `pandoc … -t html5` then cached headless Chrome `--print-to-pdf`. **Gotcha:** do *not* pass
  `--user-data-dir` (Chrome hangs). CSS at `/tmp/mdpdf/head.html`.
- **Fetched-paper QC (the `ssd/` library):** the operator's local `doc_retrieve` service returns a content-valid
  PDF that may **not match the requested DOI** — always **title-verify** each fetch against its citation (this
  caught 8 wrong papers while building the library). Use **Crossref** for correct DOIs; for pre-2010 SETAC /
  Allen-Press papers the **Wiley `10.1002/etc.5620…`** DOI often works when the `10.1897` one mismatches. OCR
  scanned PDFs with `ocrmypdf`.
- **Careful representation:** when a claim rests on a person/biographical detail, verify before asserting (e.g.
  the popular "Delaney's wife had cancer" motivation is **uncorroborated** in the authoritative histories —
  dropped). Treat the Delaney/Dingell contrast as **emblems of two philosophies**, not a personal feud.