330 FEMA numbers cover more than one record, and 64.2% of those groups contain more than one CAS

About a 4 minute read.

Last time I ran into (E)-β-damascone and β-damascone sharing FEMA 3243 while their storage instructions were opposite and their substantivity differed by 38%.

That suggested a probe: sweep the database for FEMA numbers covering more than one record, and every place the fields disagree falls out at once.

What the sweep found

Records with a FEMA number	3,617
Distinct FEMA numbers	2,968
Numbers covering more than one record	330 (11.1%)
Records involved	979

Group sizes: 219 groups of two, 50 of three, 24 of four… and the longest group holds 20.

And the internal disagreement runs higher than I expected:

	Groups	Share of duplicate groups
More than one CAS inside the group	212	64.2%
More than one odour string inside the group	203	61.5%
Every record in the group marked natural	227	68.8%

The group of twenty

FEMA 2825 is sweet orange peel oil. All 20 records share CAS 8008-57-9:

peel oil (99 suppliers), folded (122), c.p. Brazil (31), c.p. Valencia (18), c.p. Florida (16), c.p. California (11), decolorized (10), folded Valencia (10), c.p. Spain (8), c.p. Italy (8), folded Florida (6), c.p. West Indian (4), folded Brazil (4), sesquiterpeneless (4), c.p. China (2), c.p. Egypt (2), folded California (2), folded Israel (2), c.p. French Guinea (1)…

And one "sweet petitgrain oil."

Petitgrain is distilled from leaves and twigs, not peel. That is a different part of the same tree, filed under the peel oil's number. (It has zero suppliers, so you will not meet it in practice, but it shows where this number's boundary sits.)

Peppermint oil (FEMA 2848) runs to 18 records across America, India, Yakima, China, Willamette, Idaho, England, Montana, France, Morocco, Hungary, Brazil, Russia, Mongolia and Tasmania.

But the group that matters most has only four

FEMA 2665 is menthol:

Suppliers	Name	CAS
114	l-menthol	2216-51-5
38	dl-menthol	89-78-1
9	d-menthol	15356-60-2
5	(±)-menthol	1490-04-6

Four CAS numbers, four stereochemistries, one FEMA number.

And these four are not interchangeable — as the menthol article covers, the cooling comes from the l-isomer. d-menthol smells different and cools far less.

FEMA 2665 says "this may be used in food." It does not say which material this is.

Two species can share a number too

FEMA 2598 is jasmine absolute. Under it:

• jasmin absolute sambac (56 suppliers) — CAS 91770-14-8, which is Jasminum sambac
• jasmin absolute (from chassis) (46), Egypt (35), India (34), Italy (4) — CAS 8022-96-6, which is J. grandiflorum

Two species, two CAS numbers, one FEMA number. And nobody would confuse their smells.

Same with FEMA 3119, ylang ylang: ylang ylang flower oil (93 suppliers, CAS 8006-81-3) alongside grades I, II and III (42/30/45, CAS 68606-83-7). I, II and III are successive distillation fractions — materially different products.

So what is a FEMA number

It is a number for a flavour use, not an identifier for a material.

It answers "may this go into food, and in what capacity," not "what is in this bottle." At the scale of that question, twenty origins of orange peel oil really are one thing.

The problem is only in using it for something else.

Four lines of practice

1. Do not match materials by FEMA number. Two records sharing one can be different species, different stereochemistry, different plant parts.
2. Do not lean entirely on CAS either. 64.2% of duplicate groups hold more than one CAS, but the reverse happens too — those twenty orange peel oils all share one.
3. What identifies a material is the whole string: name plus origin, processing and specification. This is also where the "coarse natural name versus specific one" gap in the corpus matching article comes from.
4. When two records' numbers disagree, first confirm they are the same thing. That was last time's first question; this article turns it into a probe you can run across the database.

What this doesn't establish

• I did not check FEMA's own rules. "One number can span several CAS numbers" is an observation from the data, not something confirmed against FEMA documentation.
• I did not verify "sweet petitgrain oil sits under 2825" in FEMA's database. It may be a transcription issue here rather than FEMA's classification.
• The 61.5% odour-string difference is string comparison only. Different strings do not mean materially different smells; word order or vocabulary could account for it.
• I did not measure how severe the contradictions are within a group, only whether differences exist.
• The sensory difference between d- and l-menthol is prior knowledge, not something measured from this database.
• 330 only covers the 3,617 records that carry a FEMA number. The other forty thousand have none, and this probe cannot see them.