Verification: How the Materia Medica Was Checked

The sources

Every citation in the database was resolved against the record it points to: Crossref for anything with a DOI, PubMed for the medical literature, and the agency’s own page for a monograph or a listing. A citation that resolves is marked beside the source on its entry, with the DOI or PubMed record it resolved to. Primary texts (Pliny, Dioscorides, Culpeper, the pharmacopoeias) and standard reference works are marked as such; a book that no index carries is marked as not indexed, which is a limit of the indexes, not a doubt about the book.

Status	Sources
Read in full	290
Read in part	176
Abstract read	82
Verified against its record	544
Matched to its record and reviewed	10
Primary text	261
Standard reference work	124
Book, seen in reviews	145
Book, not indexed	154
Not indexed	11
Not yet checked	36
All distinct sources	1833

Where the press has since read a source itself, the mark beside the citation says so instead: read in full, read in part, or its abstract read, with a link to the copy that was read.

321 of the journal sources have their abstract on file. Club Members and Founding Readers read each abstract beneath its citation on the entry, so the claim and its source sit together.

The entries

Three passes were made over the 413 entries that carry a pharmacognosy block, and the dated and regulatory claims of all 476:

Whole-entry reading. Every sentence of every entry read against the abstracts of the sources it cites, with a lookup for each claim that could not be placed: fifty entries drawn at random first, then the remaining 363.
Regulatory and standing claims. Every sentence naming a monograph, a list, an approval, a ban, or a treaty (318 of them) checked against the authority itself: the EMA’s herbal pages, the WHO Model List, FDA monographs, the CITES appendices, Commission E, and the French decree of 2008.
Dated claims. Every sentence carrying a year, a century, or a dated attribution (423 of them) read against the primary record, with the doubtful ones looked up.

Before those, fifty cited claims drawn at random were compared sentence by sentence with the abstract of the source each cites. Most corrections fell in incidental history and regulatory detail; a few were to mechanisms, where an entry had named the wrong target or contradicted its own source.

The older authorities, read in full

The notes from Pliny and Culpeper, from Folkard, Thiselton-Dyer, and Grieve, and from the Catholic Encyclopedia were written from the texts themselves, fetched and read in full, not from memory of them. 183 entries carry such a note. Every quotation in them was checked mechanically, word for word, against the stored text, an elided quotation segment by segment, and corrected where it had drifted.

The sources read, not merely found

A record that a source exists is not a reading of it. On October 3, 2026 the press began reading every source that had only been found in an index or seen in a catalogue, and checking each sentence of the entry against what was read. Of the 444 entries that cite such a source, 135 have been read so far; the rest follow in the same way. In those entries 551 sources were read, in full, in part, or in their abstract; 129 entries were changed, with 869 sentences rewritten to what the source says or cut, 370 readable sources added, and 95 sources that no one could read dropped. Every passage quoted in that reading was then confirmed in a fresh copy of the text, fetched independently of the reader’s own (for 4 articles that PubMed Central would not release to a program, in the reader’s saved copy, after the archive’s own record confirmed it was that article); a rewrite that could not be so confirmed was cut rather than kept.

The mechanisms, against ChEMBL

Every constituent was looked up in ChEMBL (ChEMBL 37), the European Bioinformatics Institute’s database of bioactive molecules, and 423 link to their record; a record whose structure disagreed with PubChem’s was not linked. The mechanism lines that name a molecular target, in 71 entries, were read beside ChEMBL’s curated mechanism or its measured activity, and each entry shows the result under “Checked against ChEMBL,” with the value and its source paper, or says that nothing has been measured. Three mechanisms were corrected, and are listed below.

The corrections

46 corrections to 42 entries. Each is also recorded on its entry under Verification.

Agarwood: the claim that the Egyptians used it in embalming has no record; removed.
Basil: the estragole finding is the 2001 European opinion (genotoxic, no safe level), not a later date.
Benzoin: said the FDA lists it as a skin protectant. The FDA's 2003 final monograph expressly declined to include compound benzoin tincture.
Bezoar stone: (and the antidote reading): the Scripps test of bezoars against arsenic was in the late 1960s, not a single stated year; Paré's test on the cook was 1567 (three places).; Arrhenius and Benson's laboratory finding now dated to the late 1960s.
Black pepper: Pliny's fifteen denarii was long pepper; black was four.
Camphor: said camphor stimulates the cold receptors of the skin. ChEMBL records it as a TRPV3 agonist and, only at 400 to 660 µM, a TRPA1 blocker, with no cold-receptor record; reworded.
Cinnamon: "Egyptians had it by 2000 BC" replaced with Herodotus's embalming report; Pliny's price given as a range rather than a single 1,500 denarii.
Comfrey: "forbidden in the United States (2001) and Germany" overstated: the FDA advised manufacturers to withdraw it, and Germany allows external products only.
Crabs’ eyes: said they are still sold in the markets of Prague. No source; removed. Also "London Pharmacopoeia into the nineteenth century" softened to the eighteenth.
Cyprus powder: said atranol and chloratranol were "limited to parts per million since 2017." They are prohibited in cosmetics from 2019 under Regulation 2017/1410.
Deadly fibrecap: now "one of the few European mushrooms whose poison has a true antidote," not "the only."
Deadly nightshade: the claim that Livia and Agrippina used belladonna replaced: later writers guessed it; the Roman sources name figs and mushrooms.
Dover’s powder: "in the BP until the 1980s" softened to "into the twentieth century," unverified.
Epsom salt: the Magpie trial (2002) beat placebo; the Collaborative Eclampsia Trial (1995) beat diazepam and phenytoin. The entry had merged them.
Fig: "the EMA-listed Lenitive electuary tradition" was a confusion; the EMA lists no such thing. Now the London Pharmacopoeia's lenitive electuary.
Glauber’s salt: Glauber's spring was near Vienna, not in Hungary.
Hemp: said CBD "acts elsewhere" than THC. ChEMBL curates it as a negative allosteric modulator of CB1 and records it at GPR55 and TRPV1; the mechanism now says so.
Houndstongue: "prohibited across Europe" softened to warned against, with the EU restriction of pyrrolizidine alkaloids named.
Juniper: the kidney-irritation finding is Schilcher and Heil (1994), not "a review in 2006."
Koji mould: said the domesticated mould had lost the aflatoxin genes; the genome paper shows the cluster present but non-functional. Corrected.; a dangling fragment from an earlier edit restored to a full sentence.
Man on horseback: the Italian ban could not be verified; the French decree of 2005 stands.
Morel: "first attempts in Paris caves in 1901" replaced with the documented first French attempts of the 1880s (Roze, 1882).
Palm of Christ: said castor oil is on the WHO essential list. It is not; the list carries senna and lactulose. The EMA monograph (2016) is real and now stands alone.
Pennyroyal: said that since 1994 US products must carry a pregnancy warning. No such requirement exists; pennyroyal is sold as an unregulated supplement. Removed from the modern note and the standing.; the Royal Society report of 1665 was not of pennyroyal against rattlesnake bite. It describes a Virginian plant with leaves like pennyroyal (the American dittany) whose bruised leaves, held to the snake's face, were said to kill it. Reworded.
Peony: the "small uncontrolled study (2021)" is now named (Zangooei Pourfard et al., Epilepsy Research 176) and cited.
Quicksilver: the Minamata Convention phases dental amalgam down, not out.; Almadén mined until 2002, not later.
Reishi: dated the Bencao gangmu to 1596 while the red yeast rice entry dated it 1578. Now "completed 1578" in both.
Roquefort mould: said the fungus is on the EU qualified-presumption-of-safety list. It is not; EFSA excludes filamentous fungi from that list (while finding no evidence the cheese harms health).
Rosemary: the elderly memory trial used 750 mg, not "a teaspoon," and 6 g impaired.
Saffron milk cap: the fresco is at Herculaneum (as the lore says), not Pompeii.
Sage: Charlemagne's Capitulare de villis lists sage fifth, after lily, rose, fenugreek, and costmary, not "first."
Salt: the goitre programme of 1922 was Swiss, not "the exorcists' own country."
Saltpetre: said 5 per cent potassium nitrate toothpaste "quiets the nerve" and is FDA-approved; the cited review (Orchardson and Gillam, 2006) calls the evidence variable. Softened.
Snapdragon: Mendel worked on peas; Antirrhinum was Erwin Baur's Mendelian plant after 1900. Reworded.
Tamarind: said it is "in the European monographs for constipation." There is no EMA tamarind monograph; removed.
Troche of agaric: said agaricic acid inhibits citrate synthase. It inhibits the mitochondrial citrate carrier and the adenine nucleotide translocase.
Truffle: androstenol was found in the black truffle (Claus, 1981), not the white, and the entry's own cited source (Talou, 1990) showed pigs ignore the pure steroid and follow dimethyl sulfide. The mechanism now says so.
Turpentine: the violet smell is of the urine, not the breath.
Vinegar: acetic acid is not on the WHO essential medicines list; the WHO recommends it for visual cervical screening. Reworded.
Viper: "kills perhaps one in a hundred untreated" replaced with the record: fourteen deaths in Britain since 1876, the last in 1975.
White willow: said salicylic acid inhibits the cyclo-oxygenases as aspirin does. In the DrugMatrix screen it inhibits COX-1 and COX-2 by less than half at 10 µM; the mechanism now calls the inhibition weak.
Winter-grain mould: "killed a tenth of the population of some districts" checked and kept (reported as 10 per cent of the population in some localities).

What this does and does not mean

A verified source is one whose record was found and whose abstract, where there is one, was read against the sentence that cites it. It means the entry says what its source says. It does not make the source right, and it does not make the entry medical advice: the Materia Medica describes what the old medicine used and what the modern literature makes of it. The full records of the checks, with every citation’s status, are kept with the database’s source code.

Historical description, not medical advice.