Half the perfumery palette is declared for food use as well

A perfumer's palette and a flavourist's palette feel like different professions' toolkits. In the data they overlap more than half the time.

Of the 4,620 materials here that a buyer can actually reach, 64.8% carry at least one flavour-register number, and 2,489 are declared for flavour and fragrance use both.

Register	Materials	Share
FEMA	2,627	56.9%
FLAVIS (EU)	1,919	41.5%
JECFA	1,857	40.2%
CoE	1,783	38.6%
At least one	2,994	64.8%
None of the four	1,626	35.2%

These are per-material register numbers, not a mention of the register somewhere on a page. That distinction is not pedantic here: the previous version of this field was built by searching the whole page text for the word "JECFA", which tagged 100% of materials, and it was rebuilt from the structured columns before this article was written.

One more distinction the numbers do not make on their own: carrying an identifier is not the same as currently being listed. A withdrawn FEMA number survives in the source data with a -VOID suffix — musk ambrette is 2758-VOID, styrene 3233-VOID — and reading those as "listed" states the opposite of what happened.

FLAVIS is worse, because a deleted FL number looks exactly like a current one. Commission Regulation (EU) 2024/234 removed eight substances from the Union list for want of data ruling out consumer risk, and all eight are here, still carrying their numbers. They are now marked as deleted, along with FL 05.100 from an earlier removal — but that list is certainly incomplete, because there is nothing in the data to detect the rest with.

So treat every figure below as "has a number". It is not "is authorised today", and only the register itself can tell you which.

The third without a number is not a mystery

The obvious reading of "no register number" is "not a food material", and for once the obvious reading is right — the source data says so directly. Every material here carries a declared use, and register coverage tracks it almost exactly:

Declared use	Materials	Carry a number
Flavouring agents	700	90%
Flavour and fragrance agents	2,489	82%
Cosmetic, flavour and fragrance	158	53%
Natural substances and extractives	86	8%
Cosmetic and fragrance agents	55	7%
Fragrance agents	849	3%

A material declared for fragrance only has a one-in-thirty chance of carrying a flavour-register number. One declared as a flavouring agent has a nine-in-ten chance. The registers are doing exactly what they say on the tin, and the third of the palette without a number is mostly the third that was never a food material.

So the finding is not the 35% without. It is the 2,489 with both. More than half the accessible palette is declared for flavour and fragrance use, which is the sense in which these are one palette rather than two.

Two things this does not mean. It does not mean a botanical extract cannot be registered — 534 of the 1,196 oils, absolutes and extracts here carry a FEMA number, including patchouli oil at FEMA 2838 and clove bud oil at 2323. And it does not mean the unnumbered group is all naturals: only 39% of it is.

And then there is coumarin

Coumarin is the interesting exception, because it is a flavour material by any ordinary description and carries none of the four numbers.

The United States is explicit about why:

Food containing any added coumarin as such or as a constituent of tonka beans or tonka extract is deemed to be adulterated under the act, based upon an order published in the Federal Register of March 5, 1954.

That is 21 CFR 189.130 in full. Seventy-two years later coumarin has 74 suppliers listed here and remains standard in the tonka–hay–new-mown-grass end of perfumery.

Europe does not authorise it as an added flavouring either. Regulation (EC) No 1334/2008 puts coumarin in Annex III Part A — substances not to be added as such to food — and separately, in Part B, sets maximum levels for it where it arrives naturally inside a flavouring or a flavouring ingredient: 50 mg/kg in traditional or seasonal bakery labelled as containing cinnamon, 20 in breakfast cereals, 15 in other fine bakery ware, 5 in desserts. Those are ceilings on specific named categories of compound food, not a general permission for tonka.

The remaining two columns have their own reasons rather than borrowed ones. JECFA did evaluate coumarin, in 1981, concluding that no ADI could be established — an assessment exists, it simply is not the flavour-number listing this data records.

Which is the useful lesson in the whole table. A number here records that a particular body assigned an identifier for flavour use, in one system, at some point. It is not a property of the material, it is not a current authorisation, and it is not a safety clearance for anything.

What to do with it

• A number is not permission. It tells you an identifier exists in one system. It does not carry the conditions of use, it does not tell you the entry is current, and nothing on this page should be treated as saying a material is safe to eat. The register is the place to check that, not this table.
• If it goes on skin, the flavour registers are the wrong shelf entirely. That is IFRA's territory, and the limits work differently from how people expect.
• If you are sourcing, expect the overlap. More than half of what you are buying is declared for food use as well, which is part of why the same material shows up at wildly different prices and specifications — food-grade and fragrance-grade are different specifications of the same name.

One comparison this article deliberately does not make

Materials with a register number average 18.6 suppliers; those without average 7.9. It is tempting to read that as registers tracking the commodity end of the market.

Do not read it that way here. The two groups differ in what they are for — one is mostly declared for food use and the other mostly is not — and anything that follows from being a food material, including how many houses supply it, moves with that rather than with the register. Whatever is underneath, this data cannot separate it, and an article on this site was withdrawn this week for making exactly that inference from exactly that kind of pairing.