Beginner perfumer · 98
Fragrance splits the flower, flavour splits the fruit — one apple tree, two vocabularies
· 14 min read
The potential uses field's tags come in two halves, FR for fragrance and FL for flavour, and their shapes are nothing alike. FR has 522 tag kinds across 72,233 occurrences, averaging 138 each. FL has 590 kinds but only 11,132 occurrences, averaging 19, with a median of 3. Only 119 tags appear on both sides — 403 are FR-only and 471 FL-only. The difference is not just volume: the top FR-only tags are woody, balsam, amber, fern, oriental, chypre, which is the entire classical perfumery vocabulary, while the top FL-only tags are tropical, tea, nut, chocolate, meat, tomato, cheese. The cleanest comparison is the apple: FL carries 25 apple tags and distinguishes Fuji, Gala, Golden Delicious, Granny Smith, McIntosh and Red Delicious; FR carries six, one of which is apple blossom at 237 uses — while flower tags on the FL side occur nine times in total.
About a 5 minute read.
I wrote about the fragrance half of potential_uses earlier. This is the flavour half, and what shows up when you put the two side by side.
The two halves look nothing alike
Across the 7,807 records with this field, tags split into FR (fragrance use) and FL (flavour use):
| Tag kinds | Occurrences | Mean per tag | Median | |
|---|---|---|---|---|
| FR | 522 | 72,233 | 138.4 | 96 |
| FL | 590 | 11,132 | 18.9 | 3 |
FL has more kinds than FR and a seventh of the occurrences.
The tails differ too: 31.5% of FL tags occur exactly once, against 19.7% of FR tags.
Only 119 tags appear on both sides — 403 are FR-only, 471 FL-only.
The exclusive lists say it plainly
FR-only, most frequent:
woody (1,186), floral (1,170), balsam (887), amber (798), fern (752), oriental (698), fruit (679), chypre (506), aldehydic (497), bouquet (492), moss (476), lavender (474)
That is the classical perfumery vocabulary entire. "Fern" is fougère and chypre is chypre — neither word appears once on the flavour side.
FL-only, most frequent:
tropical (428), tea (325), nut (313), chocolate cocoa (256), meat (248), tomato (211), caramel (199), dairy (186), beverage (163), vegetable (160), cheese (147)
Those are the names of foods.
Fragrance names the abstract categories it invented; flavour names things. (I read both top thirties by hand: 14 of the FR tags are abstract families, and 2 of the FL ones.)
And the cleanest comparison is the apple
Same word at the front, two entirely different ways of subdividing:
FL carries 25 apple tags:
apple (53), apple green apple (16), apple red apple (10), apple cider (3), apple delicious apple (3), apple spice (3), apple baked apple (2), apple cinnamon (2), apple pie (2), apple sour apple (2), and one occurrence each of apple fuji apple, apple gala apple, apple golden delicious apple, apple granny smith apple, apple mcintosh apple, apple red delicious apple, apple juice, apple sauce, apple butter, apple spiced apple, apple enhancer, apple fractions, apple apricot, apple cranberry, apple baked spice apple
Six named cultivars.
FR carries six: apple (484), apple blossom (237), apple green apple (152), apple red apple (84), apple candy apple (1), apple crisp (1)
Note the second one.
Which side has the flowers
| Tags containing blossom or flower | Occurrences | |
|---|---|---|
| FR | 27 kinds | 3,717 |
| FL | 2 kinds | 9 |
A factor of 413.
FR's flower tags: orange blossom (357), linden flower (294), crabapple blossom (271), cherry blossom (243), apple blossom (237), elder flower (217), plum blossom (192), iris blossom (184)…
FL's flower tags, in full: orange blossom (8), cherry blossom (1).
One apple tree: fragrance buys the flower, flavour buys the fruit. And each industry subdivides its own half to the limit.
The pattern repeats
| FL subdivisions | FR subdivisions | |
|---|---|---|
| Apple | 25 | 6 |
| Tea | 11 (Darjeeling, oolong, chai, rooibos, jasmine, lemon tea…) | 2 |
| Orange | 10 | 5 |
| Cherry | 8 | 5 |
| Mango | 2 | 4 (including mango flower, 139) |
| Plum | 3 | 3 (including plum blossom, 192) |
Tea makes it especially clear: the flavour side resolves Darjeeling and oolong, while the fragrance side has green tea and orange pekoe tea, the second occurring once.
Why
Because each industry subdivides only where it can sell the distinction.
A flavour house sells "Fuji apple flavour" because a beverage company asks for it. Nobody sells a Granny Smith accord — they sell apple, or they sell apple blossom.
And the reverse holds. Nobody sells apple-blossom-flavoured juice.
This is not about which vocabulary is finer. The two answer different questions.
And the flavour side's granularity has analytical chemistry behind it. George and colleagues' 2024 Food Chemistry paper is exactly that work — stable isotope dilution with HS-SPME-GCMS quantifying pineapple's key aroma volatiles in one run:
The method was applied to identify the key aroma volatile compounds produced by popular pineapple cultivars such as 'Aus Carnival', 'Aus Festival', 'Aus Jubilee', 'Aus Smooth (Smooth Cayenne)' …
Five named pineapple cultivars, 26 target compounds. The same thing as those six apple cultivars in two different forms — one a vocabulary, the other a mass spectrometer.
In practice
- Do not search for flavour uses with perfumery words. Fougère does not exist on the FL side.
- Do not expect FL tags to find you floral material. That side has nine flower tag occurrences in total.
- The 119 shared tags are the bridge — concepts that mean something to both industries (fruity, citrus, green and their kind).
- A material carrying many tags on both sides is worth a look: it occupies the uncommon position of being mainstream in both.
What this doesn't establish
- Tags come from splitting strings on
;;and|, so irregularly written entries are missed. - "14 of FR's top thirty are abstract families" is my own hand classification, and the line is arguable (is "fruit" abstract or concrete?).
- I did not check where these tags come from. They are the database's own classification, not any authority's list.
- "They subdivide where it sells" is my interpretation; the database does not explain its scheme.
- George 2024 is a methods paper, not about vocabularies. I cite it to show that cultivar-level differences have measured backing, not because it discusses any of this.
- Tag counts are not importance. A tag appearing 400 times may just mean one batch of records was labelled in bulk.