How I Trace a Herb's Effects to the Original Text

How I Pin a Herb's Effects Back to the Original Classical Text

I keep saying MateriaTrace is "source-traceable" — every field can be tracked back to the original classical text. It sounds good. But what does it actually mean, and where does the work actually go?

This article is about the finicky parts. Most of them are small, even a little obsessive. But those small things are exactly what separates a real "source-traceable" claim from one that just sounds nice.

Grunt work #1: pin each field to its exact spot in the text

The most basic step: every field in the database isn't my summary. It's lifted word-for-word from the original text, with a record of which scroll, which clause, which passage it came from.

Take astragalus (huang qi). Its traditional functions aren't "a few words I came up with after reading." They're a verbatim quote from the Bencao Gangmu — with each line attributed to the Shennong Bencao Jing, the Mingyi Bielu, or a named physician. Every record carries a content fingerprint, so you can take the original book and confirm the text is there, exactly, at that spot.

Why go this far? Because "source-traceable" that only means "I consulted a book" is worthless — once you've consulted it and paraphrased in your own words, the reader has no way to check. Only by pinning the original text and its location does traceability become verifiable instead of a slogan.

Grunt work #2: tell apart things that look the same but aren't

Classical texts are full of traps where two things look identical but are actually different herbs. Get sloppy and you mix them up — and the moment you do, the whole chain of sourcing is a lie. Three I actually ran into:

Two herbs, one spelling. Da ji (Euphorbia pekinensis) and da ji (Cirsium japonicum) are written with different characters but romanize to the exact same "da ji" in pinyin — and they're completely different herbs. One is a toxic, drastic water-purging plant; the other cools the blood and stops bleeding. The Ten-Jujube Decoction uses the first one. Match by pinyin on autopilot and you'd link the formula to the wrong plant entirely. So this kind of thing has to be checked against the original, herb by herb, to confirm you've linked the right one.

Different parts of one entry. In the Bencao Gangmu, "trichosanthes" is a single entry — but it records several parts: the root, the fruit, the seed, each with different properties and uses. Early on I'd only captured the root (that's trichosanthes root / tian hua fen). Later, to add the fruit (gua lou shi), I had to pull the "fruit" passage out on its own and give it its own entry — and never let the fruit's record bleed into the root's. The root's aliases stay with the root; the fruit's properties stay with the fruit; each is its own original text.

Two formulas, one name. Classical formulas include cases where the same name isn't the same formula. In the Shanghan Lun, Ge Gen Tang is the formula given in clause 31. But there's another "Ge Gen Tang" elsewhere that's actually a variant of a different formula — and the editor Lin Yi even noted in the margin that it was "probably not the original intent." When you enter this, you have to nail down which clause and which version. Pick the wrong one and the formula's whole composition and source are wrong.

These distinctions are tedious, slow, and unglamorous. They are also the exact line between "source-traceable" and "close enough."

Grunt work #3: when the identity isn't clean, leave it blank

Sometimes the old name doesn't map cleanly onto a single modern species. There, being rigorous means showing restraint — better to leave it blank, or flag it as uncertain, than to force a clean-looking answer.

An example. Several formulas in the Jingui Yaolue use "wu tou" (aconite). The database has both wild aconite / cao wu (one species) and aconite / fu zi (the cultivated aconite system — a different species). The Jingui's "wu tou" refers to the cultivated kind — same plant as fu zi, different part. Take the lazy route and link it to the most similar-looking name, "cao wu," and you've got the wrong species.

So the call here was: identify it as belonging to the cultivated fu zi system, record the original name "wu tou" faithfully — rather than invent a new entry with no textual basis, or link to the wrong species. When the text can't support a clean conclusion, the honest move is to mark it uncertain, not to fabricate a complete-looking answer. Filling gaps from thin air is the one thing a source-traceable database can never do.

Why be this rigorous

You might ask: does it really matter? A wrong internal link, a quote that's not quite precise — no ordinary reader would even notice.

Precisely because they can't notice, it matters more.

A database that claims to be source-traceable rests its entire value on one thing: that every single line really can be traced, really does check out. The instant one spot is "close enough," or "probably this one," or filled in from memory, that premise breaks — and the reader has no way to know which spot broke. So they're left to either distrust all of it or trust it blindly. Neither is what I want.

So I'd rather be slow. Rather check herb by herb, line by line, and leave it blank when it doesn't check out, than trade "every line holds up" for "looks complete." In an era when AI can generate endless complete-looking content in seconds, "holds up" is the genuinely scarce, genuinely valuable thing.

To be clear: MateriaTrace records how the classical texts read, plus objective facts about a substance — it is not medical advice. Modern science's verdict on Chinese medicine is cautious and mixed (the NCCIH notes studies have had mixed results with no firm conclusions, and warns against using TCM to replace or delay conventional care).

See for yourself

All this grunt work lands on the pages you can click. Open astragalus and watch its functions trace line by line back to the Bencao Gangmu; or compare da ji the Euphorbia with da ji the Cirsium and see how two herbs with the same spelling are kept cleanly apart.

If this "pin everything back to the original" rigor is useful to you, see what else is in MateriaTrace on the homepage.

MateriaTrace is a source-traceable, bilingual knowledge graph of Chinese medicine. It records how the classical texts describe each entry, with every field traceable to the primary source, and makes no treatment or usage recommendations. Every classical record in this article can be checked word for word on its entry page.