A Classical Text Calls Tofu "Slightly Toxic."

A Classical Text Calls Tofu "Slightly Toxic." It Doesn't Mean What You Think.

A record that stops you short

Look up tofu in the Bencao Gangmu, and under its nature and flavor you'll find: "sweet, salty, cold, slightly toxic."

Tofu? Toxic? The tofu you eat every week?

Keep reading and it's not a fluke. Garlic is "toxic." Cucumber is "slightly toxic." Lettuce is "mildly toxic." Even grape wine gets "mildly toxic." Taken literally, Li Shizhen labeled half a produce aisle as poison.

He didn't mean that, of course. The problem is this: the word "toxic" in a 16th-century Chinese herbal and the word "toxic" today are two completely different systems. Miss that, and you'll badly misread a huge amount of what the classics actually say.

The old "toxic": a scale of how strong a substance is

In classical Chinese pharmacology, "toxic" often doesn't mean "a harmful substance that can poison you." It means how strong and forceful a substance is — how potent it acts, how carefully you'd need to handle it.

So the old texts graded things along a scale: greatly toxic, toxic, slightly toxic, mildly toxic, down to "non-toxic." Think of it less as a modern toxicology rating and more as a dial for "how powerful is this, and how much caution does it call for."

On that dial:

• Tofu at "slightly toxic" and garlic at "toxic" are really saying these run strong, that overdoing them might not sit well — a note about character and moderation, not "this will poison you."
• Aconite (fu zi) at "greatly toxic" is a different story entirely — a marker that this is extremely forceful, requiring processing to temper it and strict limits on use.

The same word, "toxic," carries completely different weight on tofu and on aconite. That's exactly why you can't lift the old "toxic" out of its context and read it as the modern one.

The modern "toxic": chemistry and pharmacology

The modern sense of "toxic" runs on a different standard entirely — it points to actual harm at the level of chemistry and pharmacology: a known poisonous compound, damage to the liver or kidneys, a specific toxin.

Apply that to Chinese herbs and you get three different situations.

One: they line up. Some substances the old texts called "greatly toxic" really do contain genuine toxic compounds. Aconite (fu zi) contains aconitine; croton seed (ba dou) is a violent purgative. Here the classical "strength" grade and the modern chemistry agree. The ancients flagged it by experience; modern science found the reason.

Two: they diverge — old says toxic, modern doesn't. Tofu and garlic got a "slightly toxic" / "toxic" label from dietary experience, but chemically they aren't harmful substances. Here the old "toxic" is a note on character, not a chemical hazard.

Three: they diverge the other way — old says non-toxic, the substance isn't. It runs backward too. Cinnabar (dan sha) is recorded as "non-toxic" in the classics — but cinnabar is mercuric sulfide. It contains mercury, which is a modern, established fact. The ancients had no concept of heavy metals, so of course they couldn't flag it.

Put together, these three cases make one point: the old and the modern systems for "toxic" sometimes match and sometimes don't. Treat one as the other, and you'll either panic over nothing or miss a real problem.

How MateriaTrace handles this: report the text, don't rewrite it as a modern verdict

Once you see the gap, MateriaTrace's rule about toxicity makes sense: we record exactly how the classical text labels something, and we don't rewrite it into a modern conclusion on the author's behalf.

In practice:

• If the text says "slightly toxic," we record "slightly toxic," and mark that this is the classical grading — even for a substance that looks harmless today. The tofu page tells you the Bencao Gangmu called it "slightly toxic." It does not tell you "tofu is poison, don't eat it."
• Where there's a plain modern fact about the substance itself (cinnabar contains mercury, minium is a lead oxide), we add one neutral note about its composition — that's chemistry, not a claim about effect or use.
• What we don't do: translate the old "toxic" into the modern "toxic" for you, and certainly not tell you what's safe to eat, what isn't, or how to reduce toxicity. That's medical guidance, and it's not our job.

Why so restrained? Because the moment we "translate" the old into a modern verdict, we've slipped our own judgment into the classical record — and the reader can no longer tell which sentence is Li Shizhen and which is us. Keeping "what the text records" and "what's a modern fact about the substance" cleanly separate, each on its own footing, is what respecting the reader looks like.

Worth saying plainly: modern science's verdict on Chinese medicine overall is cautious and mixed. The U.S. National Center for Complementary and Integrative Health (NCCIH) states that studies have had mixed results with no firm conclusions, and warns against using TCM to replace or delay conventional care. MateriaTrace records what the historical texts say; it is not medical advice.

Read "toxic" right, and you can read the classics

Next time you see a kitchen-staple food flagged as "toxic" in an old herbal, don't jump. Ask one question: is this the old sense — "this acts strongly" — or the modern one — "this is chemically harmful"? The answers are often not the same.

That's the whole point of being source-traceable. When every instance of "toxic" clicks back to the original text — which book, which line — with the known modern facts about the substance noted alongside, you can tell the two apart yourself, instead of being scared or misled by a vague label.

Want to see it? Open aconite and look at how the Bencao Gangmu marks its "greatly toxic"; then open tofu and see what the same word weighs on a kitchen staple. Read them side by side and this whole article clicks into place.

If this "everything clicks back to the original" 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 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.