The Eighteen Antagonisms and Nineteen Fears: What the Classics Recorded About Combining Herbs
Two mnemonic verses, centuries apart, name pairs and groups of herbs the classical tradition recorded as antagonistic — 十八反 (eighteen antagonisms) and 十九畏 (nineteen fears). Each verse is pinned to its own source. Neither list stayed fixed — even the count named in "eighteen" grew across later books. And the very authors who recorded the rule used the "forbidden" pairs in their own formulas. Here is the classical record, kept exactly as written, tension and all.
Two Verses of Herbs That "Fight" Each Other — Recorded, Named, and Then Used Anyway
Two verses, memorized for centuries
Chinese materia medica has a seven-part classical theory of how substances behave together — danxing (used alone), xiangxu and xiangshi (mutual reinforcement), xiangwu (mutual suppression), xiangwei (mutual restraint), xiangsha (mutual neutralizing), and xiangfan (mutual antagonism). The Shennong Bencao Jing's preface named all seven; Li Shizhen restated them in the Bencao Gangmu's own opening chapter. Two of those seven categories were eventually distilled into verses meant to be memorized outright: the eighteen antagonisms (十八反, shi ba fan) and the nineteen fears (十九畏, shi jiu wei). This article records both verses exactly as the classical tradition preserved them — where each first appears, which herbs each names, and a tension the tradition itself never resolved. It is a record of what old books say about combining herbs with each other. It says nothing about combining any herb with a modern medicine, and it recommends nothing to anyone.
The eighteen antagonisms (十八反)
The earliest surviving version of the mnemonic verse appears in the Jin-dynasty physician Zhang Zihe's Ru Men Shi Qin ("Confucians' Duties to Their Parents," a title referring to the duty of a learned person to understand medicine):
本草明言十八反,半蔞貝蘞及攻烏,藻戟遂芫俱戰草,諸參辛芍叛藜蘆。
"The materia medica plainly states the eighteen antagonisms: ban [xia], [gua] lou, bei [mu], and [bai] lian attack wu [tou]; [hai] zao, [da] ji, [gan] sui, and [yuan] hua all make war on [gan] cao; the various shen, [xi] xin, and shao [yao] turn against li lu."
Spelled out, the verse names three groups:
- Wu tou (aconite — chuan wu/cao wu/fu zi) antagonizes: bei mu (fritillaria, all varieties), gua lou (trichosanthes fruit, peel, and seed) and its root tian hua fen, ban xia, bai ji, and bai lian.
- Gan cao (licorice) antagonizes: gan sui, da ji (both the jing and hong varieties), hai zao, and yuan hua.
- Li lu (veratrum) antagonizes: ren shen, dan shen, xuan shen, sha shen (nan and bei varieties), ku shen, xi xin, and shao yao (both bai shao and chi shao).
Modern Zhong Yao Xue textbooks add xi yang shen (American ginseng) and dang shen to the li lu group — herbs that entered Chinese materia medica after trade routes brought them in, centuries after the Bencao Gangmu was compiled, so neither appears in the 1596 text itself. That is a textbook's own later expansion of the classical list, kept visible here rather than folded silently into "the classics said."
The name itself outlived any fixed count. A tenth-century scholar of the Later Shu, Han Baosheng, revising the Shu Bencao, was the first to tally the seven relations and record "eighteen antagonistic drugs" as a number. By the Zhenglei Bencao the list had grown to 24 antagonistic pairs; the Bencao Gangmu itself records 36. The name "eighteen antagonisms" was kept regardless of how the underlying list grew — it had already stopped meaning a fixed count of eighteen.
The nineteen fears (十九畏)
The earliest surviving version of this second verse appears in the Ming-dynasty physician Liu Chun's Yi Jing Xiao Xue ("Elementary Studies for Medicine"):
硫黃原是火中精,朴硝一見便相爭,水銀莫與砒霜見,狼毒最怕密陀僧,巴豆性烈最為上,偏與牽牛不順情,丁香莫與鬱金見,牙硝難合京三棱,川烏、草烏不順犀,人參最怕五靈脂,官桂善能調冷氣,若逢石脂便相欺,大凡修合看順逆,炮爁炙煿莫相依。
"Sulfur is the essence born of fire, yet meets mirabilite and the two contend at once; mercury must never meet arsenic; langdu fears litharge above all; croton seed's nature is the fiercest, yet it will not settle beside pharbitis seed; clove must never meet curcuma; refined mirabilite will not combine with sparganium; chuan wu and cao wu do not agree with rhinoceros horn; ginseng fears flying-squirrel droppings above all; cinnamon bark can regulate cold qi well, yet meets its match in the mineral clay; in compounding, always watch what agrees and what does not — processing by fire must never bring the wrong two together."
Nine pairings, nineteen named substances in total:
- Sulfur fears mirabilite
- Mercury fears arsenic
- Langdu fears litharge
- Croton seed fears pharbitis seed
- Clove fears curcuma
- Chuan wu and cao wu fear rhinoceros horn
- Refined mirabilite (a further-processed form of mirabilite) fears sparganium
- Cinnamon bark fears the mineral clay (chi shi zhi)
- Ginseng fears flying-squirrel droppings (wu ling zhi)
Two entries are worth a second look. Mirabilite appears in the verse twice under two different names — as the crude form (pu xiao, pairing 1) and again as the more refined form (ya xiao, pairing 7) — because in classical processing they were treated as different-enough substances to fear different partners, even though both derive from the same mineral. And the same rhinoceros horn recorded in our earlier piece on the classical record and the ban reappears here, in an entirely different classical role: not as a substance used on its own, but as one aconite was recorded to "fear."
The tension the tradition never resolved
The classical record on whether fan (antagonistic) pairs may ever be used together is not one voice — it is an argument that runs for centuries, on both sides sourced.
Against combining them: the Shennong Bencao Jing's preface itself instructs, "do not use those that are mutually restraining or mutually antagonistic." Tao Hongjing's Bencao Jing Jizhu is blunter: "where they are mutually antagonistic, the two are as enemies to each other and must not be combined." Sun Simiao went further still: "herbs and minerals that are mutually antagonistic will derange a person — the force is greater than a blade."
And yet the record of formulas that combine exactly these "forbidden" pairs runs just as deep, in some of the same authors' own hands. Zhang Zhongjing's Jinkui Yaolue pairs gan sui with gan cao in Gan Sui Ban Xia Tang, and pairs wu tou with ban xia in Chi Wan. Sun Simiao's own Qian Jin Yi Fang uses wu tou together with ban xia, gua lou, bai ji, and bai lian in several formulas. Most pointed of all: Zhang Zihe — the very physician whose verse gave the eighteen antagonisms their name — records hai zao together with gan cao in his own formula, Tong Qi Wan, in the same book that names the two as antagonists. Li Shizhen, compiling all of this centuries later, offered no tidy resolution either: "where mutually-restraining and mutually-antagonistic substances are used together, it is the way of the hegemon — there is the constant and there is the exception, and it is for the one who uses them to recognize which is which."
We keep that tension visible rather than resolve it. The verses that name the antagonisms are real, sourced, and quoted above exactly as recorded. So are the formulas, by some of the same classical authors, that combine the pairs those verses name. Neither side of that argument is ours to settle.
How MateriaTrace handles it
Two verses, each pinned to the book that first recorded it. A list whose name stayed fixed while its contents grew across three later texts. A rule some of its own recorders didn't follow in their own formulas. We record all of it as what it is — a classical debate, never resolved within the tradition itself — rather than flattening it into a single instruction.
Worth stating plainly: this article is a transcription of a classical, herb-to-herb theoretical framework. It is not medical advice, it names no modern drug interactions, and it recommends nothing about what to combine or avoid. MateriaTrace records what classical sources say and makes no treatment or combination recommendation; any question about combining substances belongs with a qualified professional.
If this "every record clicks back to its own source" approach 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 classical and modern sources describe each entry — each field traceable to its primary source, contradictions kept side by side — and makes no treatment or usage recommendations. Every record in this article can be checked on its entry page.
Related entries
Look up 1,500+ source-traceable herbs?
MateriaTrace is a free, open archive — every entry is fully cited and open to browse, no account required.
Browse the archive