348 records give the dose in millilitres, and a millilitre is not a dose

5,108 records carry a toxicity figure in mg/kg. 348 carry one in ml/kg. Turning a volume into a dose requires a density, and only 126 of those 348 (36.2%) have a specific gravity on file — the other 222 cannot be converted from the database alone. Worse, 16 of them give ml/kg for something the appearance field calls a powder, crystals, or a waxy paste, where millilitres correspond to no meaningful amount fed to anything. But the field carries its own antidote: 85.1% of the ml/kg records also hold an mg/kg number in the same cell, so the two can be checked against each other. Acetophenone's cell holds 3000 mg/kg, 3200 ml/kg and 2.48 ml/kg at once — and converting shows that the 3200 is the right number with the wrong unit.

About a 4 minute read.

While writing up treemoss last time, I found this in the toxicity field: oral-rat LD50 4330 ml/kg. The unit is millilitres.

For a 250 g rat, 4,330 ml/kg works out to more than a litre. Not possible. All I could say at the time was "do not compute with this cell," so I went back to count how many records look like that.

348.

The unit distribution first

Scanning the text of all three toxicity fields and pulling the unit after each number:

Unit	Records
mg/kg	5,108
mg/m³	529
ml/kg	348
ppm	293
µg/kg	259
g/kg	3

ml/kg is the third most common unit. This is not one or two slips, it is a convention — older toxicology measured liquids by volume, because when you dose an animal with a pipette, volume is what you actually measure.

The problem is not that it is wrong. The problem is that it is incomplete.

Turning volume back into dose needs a density

X ml/kg × density = Y mg/kg. The density is right there in the specific gravity field.

But only 126 of those 348 records have one — 36.2%.

For the other 222, the database alone cannot convert it. You have to look the density up yourself, and for a natural, the density has a range of its own.

For 16 of them, millilitres do not apply at all

I pulled the records whose appearance field says solid, crystal, powder or wax with no mention of liquid at all:

Appearance	Material	Value
White granular solid	Dicyclohexyl phthalate	30 ml/kg
Colorless white solid	1-dodecanol	10 ml/kg
Cream waxy paste	PEG-8 stearate	27 ml/kg
Yellow orange semi-solid paste	Nutmeg oleoresin	10 ml/kg
Colorless to amber powder	4-tert-butyl phenol	3.2 ml/kg

"10 millilitres of a powder" is not a meaningful dose — the bulk density of a powder depends on how you pack it.

(The three treemoss records read "dark brown semi-solid," and they have no specific gravity field at all, so there is not even an entrance to the conversion.)

But the field carries its own antidote

296 of the 348 (85.1%) also hold an mg/kg number in the same cell. Same material, same route, numbers from different sources sitting side by side — which means they can be checked against each other.

Acetophenone (#281, 44 suppliers) is the cleanest case. Its toxicity_oral cell packs five oral rat figures:

Value	Source
3000 mg/kg	Smyth & Carpenter, 1944
3200 ml/kg	Jenner et al., 1964
2.48 ml/kg	Smyth et al., 1969b
900 mg/kg	Smyth & Carpenter, 1948
2700 mg/kg	1988, mammal species unspecified

Acetophenone's specific gravity is 1.025 to 1.032. Converting:

2.48 ml/kg × 1.028 ≈ 2,550 mg/kg — which agrees with the 3000 mg/kg from 1944
3200 ml/kg × 1.028 ≈ 3,290,000 mg/kg — 3.3 times the animal's body weight

So the 3200 is the right number with the wrong unit. It should read 3200 mg/kg — and that value is almost identical to the 1944 figure in the same cell.

This is not a problem with volume units. It is a transcription problem. And it was catchable only because there were other numbers in the same cell to check against.

Converted across the database, 8 numbers do not stand

Converting all 167 ml/kg values on the 126 records that have a specific gravity gives a median of 11,080 mg/kg (which itself says ml/kg tends to get used on low-toxicity things — corn oil 100 ml/kg, cottonseed oil 90 ml/kg, where the answer is essentially "you cannot kill it with this").

Eight land above 50,000 mg/kg, meaning 5% or more of the animal's body weight would have to be the substance:

Converted	Original	Material
3,280,000	3200 ml/kg	Acetophenone (44 suppliers)
178,800	200 ml/kg	Butyl acrylate (12)
65,152	64 ml/kg	PPG glyceryl ethers (five records)
62,640	60 ml/kg	Polysorbate 60 (22)

The five PPG records carry the same value, which looks like one source copied across the group. Four of them have zero suppliers.

And while we are here: the same authors disagree with themselves

Back to the acetophenone cell. Smyth & Carpenter published 3000 mg/kg in 1944 and 900 mg/kg in 1948. Same people, four years apart, 3.3× apart.

That is not an error. That is the spread inherent in an LD50. The lowest and highest in this one cell differ by 3.6×, and all of them are real.

How to read this field in practice

See ml/kg, look for the specific gravity first. No density means no number.
Sanity check after converting. Anything above 5% of body weight, suspect the unit.
The other numbers in the same cell are your control. 85% of the time you have one.
Skip ml/kg on a solid.
Do not take the lowest number in a cell as the answer. A 3× spread within one cell is normal.

What this doesn't establish

I went to none of the primary sources. Smyth 1944, Jenner 1964 and the rest are citation strings inside a field, nothing more.
"The 3200 should be mg/kg" is an inference from the other numbers in the same cell and the implausibility of the conversion, not something I confirmed in the original paper.
Unit extraction relies on a regular expression. Unusually written records may be missed, so 348 is a floor.
I took the first number in each specific gravity range. I did not handle density ranges for naturals.
"ml/kg tends to be used on low-toxicity things" is a correlation read off the median, with no test behind it.
The 16 solids were classified from the appearance string. Records reading "liquid to solid" were excluded, so that is a floor too.