Kkaennip is not shiso, and the difference has been measured

Every English-language guide to this leaf spends a paragraph on the same confusion, and then stops in the same place. Perilla and shiso are related, they say; perilla is bigger and less citrusy; do not call it sesame leaf. All true. What none of them mentions is that the difference has been measured, in a laboratory, with the two plants grown side by side — or that the most practical question about the leaf, what happens to its smell when you cook it, appears never to have been answered at all.

This article is built from published sources rather than from my own kitchen. Every claim below names the paper or the agency it comes from, and the last section lists what I looked for and could not find.

The same species, a different variety, and Korea does not grow the other one

Kkaennip is Perilla frutescens var. frutescens. Shiso is Perilla frutescens var. crispa. One species, two varieties — which is why the plants cross-breed and why the naming is such a mess in English.

The part that does not appear in English writing is what happens when you go looking for shiso in Korea. In collecting trips across the country, researchers gathered 176 perilla accessions and found var. frutescens cultivated at scale everywhere, while var. crispa appeared only as a weedy form — 20 samples, none of them under cultivation (Lee, Kim & Ohnishi, 2007). Shiso is not a Korean crop. It grows in Korea the way a volunteer tomato grows in a compost heap.

One caution about names. International taxonomic databases use var. frutescens; Korean agronomy papers routinely write var. japonica for the same plant. Both refer to the perilla grown for leaves and seeds in Korea. I have not found a Korean government page that settles which name it considers correct.

They do not smell alike, and this is the measurement that shows it

Essential-oil composition varies with where a plant grew, so comparing a Korean study to a Japanese one proves little. One study avoids that problem entirely: three perilla taxa raised together in the collection at Kaunas Botanical Garden in Lithuania, harvested at full flowering in August 2007, analysed on the same instruments (Bumblauskienė et al., 2009).

• Perilla frutescens — “Essential oil was ascertained to perillaketone chemotype. Principal compounds were found to be perillaketone and egomaketone.”
• P. frutescens var. crispa f. viridis (green shiso) — perillaldehyde 49.47 %, limonene 11.76 %, limonene oxides 9.85 %, caryophyllene oxide 7.21 %.

Different dominant molecules, not different intensities of the same one. Perilla ketone and perillaldehyde are separate compounds with separate smells, which is why substituting one leaf for the other does not work the way substituting parsley for chervil might.

Korean material gives the same answer. Analysing 80 Korean accessions, cultivated var. frutescens came out at 87.2 % perilla ketone, while the weedy var. crispa samples led with perilla aldehyde at 26.4 % (Cho et al., 2024; Sa et al., 2023). A separate study of purple and green perilla found the two compounds swapping places almost completely: perillaldehyde 31.72 % in purple against 0.60 % in green, and perilla ketone 27.50 % in green against 2.15 % in purple (Chen et al., 2022).

There is a published classification behind this. Perilla is sorted into chemotypes by whichever volatile dominates — PA for perillaldehyde, PK for perilla ketone, PL, EK, PT and several phenylpropanoid types — with a pure chemotype defined as one compound above 50 % of the total (Wei et al., 2025). The leaf Koreans eat is a PK type. The leaf used in Japanese cooking is a PA type.

“87 % perilla ketone” is not a fact until you say how it was extracted

This number gets quoted as though it were a property of the leaf. It is partly a property of the method.

Working on Korean perilla leaf, one group reported perilla ketone at 81 % by solvent-assisted flavour evaporation, 84 % by liquid–liquid continuous extraction and 95 % by hydrodistillation — the same leaf, three figures (Seo & Baek, 2009). A decade earlier, researchers at the Rural Development Administration’s National Crop Experiment Station compared three other methods and got roughly 62 % by solvent extraction, 86 % by headspace and 92 % by simultaneous distillation-extraction (Kim et al., 1999).

So the honest form of the sentence is “perilla ketone made up 95 % of the oil by hydrodistillation”, not “perilla leaf is 95 % perilla ketone”. Any article that gives you a single percentage without a method has copied it from somewhere without checking what it means.

The same thing has happened to miyeok, where “expands ten times” circulates without anyone saying ten times by weight or by volume, or which of the nine dried seaweed products is meant.

The smell is on the underside, locked in glands

Perilla’s oil is not distributed through the leaf. It sits in glandular trichomes — oil glands — and a survey counting them found the large peltate glands only on the underside: 657 on the abaxial surface and zero on the adaxial, with heads about 72 µm across against 25 µm for the smaller glands on the upper face (Plant Methods, 2023).

Two cautions, both of which matter. The paper does not state whether 657 is a per-square-millimetre density or a total across the ten fields observed, so the number should not be quoted as a density. And the plants were grown from seed bought at a herb market in Anguo, Hebei, China — not a Korean culinary cultivar.

What the glands do has been measured, on shiso rather than kkaennip. Imaging the leaf showed perillaldehyde concentrated in the oil glands with only “a small amount” in the epidermis, and crushing the leaf into a paste raised measured perillaldehyde from 0.20 mg/g to 12.6 mg/g — about sixty-three times (Taira et al., 2021). The aroma is not released by the leaf being present. It is released by the glands being broken.

That reframes the ordinary uses. Wrapping meat in a whole leaf, slicing it into ribbons, pounding it into a seasoning paste and sealing it inside pancake batter are four different mechanical events before heat is considered at all.

Nobody appears to have measured what cooking does to it

This is the part I expected to be easy and was not.

I went looking for any study quantifying the aroma of perilla leaf after pan-frying, boiling, steaming, blanching or frying in batter — volatile retention, a temperature threshold, a time course, anything. Five structured searches of the literature, with every returned title checked: a volatiles-and-heat query returned seven papers, a cooking-methods query returned twenty-seven, a drying query returned thirty-seven, and two Korean-language searches returned none. Not one measured a cooked leaf.

English-language food writing does not fill the gap either. It does not raise the question. One guide notes that “even the most pungent varieties lose much of their aroma when dried” and tells you to use the leaves fresh; heat is not discussed.

The nearest published measurement is not fresh leaf at all. Perilla leaves were dried at 60 °C (140 °F), ground, and roasted at 150 °C (300 °F) for up to eight minutes (Lee et al., 2019). Two compounds show the shape of what heat does:

	0 min	2 min	4 min	6 min	8 min
Methyl benzoate (µg/100 g)	88.08	112.46	203.10	122.67	21.31
Limonene (µg/100 g)	54.66	58.18	137.24	60.92	12.56

Volatiles rise to roughly two and a half times their starting level at four minutes, then collapse to under a quarter by eight. Heat brings the smell out before it drives it off.

Read that with care, though. Neither perilla ketone nor perillaldehyde appears anywhere in that study’s results — the two compounds that define the leaf’s smell were not detected at all, most likely lost during the 60 °C drying step, although the authors do not discuss it. The paper also contradicts itself on the eight-minute limonene figure, printing 60.92 in the text and 12.56 in the table. I have used the table.

So the useful statement is a negative one: the question of what happens to kkaennip in a hot pan is open, and the confident answers circulating in recipes are not resting on anything published.

The storage advice in English is backwards

Here a Korean government source contradicts the English consensus directly, and it is the most practically useful finding in this article.

English guides say to keep perilla leaves in the coldest part of the fridge — the crisper drawer — and give about a week. The Rural Development Administration’s Postharvest Technology Division says the cold is the problem:

“깻잎을 냉장고 안쪽(1∼4℃)에 보관할 경우 잎에 검은 반점이 생기는 저온 장해와 수분 손실로 인한 시듦 현상이 발생하기 쉽습니다.”

Stored at 1–4 °C (34–39 °F), the leaves develop chilling injury — those black spots — and wilt from moisture loss. The agency’s own figure is that the leaf is about 90 % water and sensitive to cold after harvest. Its recommendation is the door shelf at around 5–6 °C (41–43 °F), where respiration slows and yellowing is delayed. With a bag it developed for the purpose — 0.1 mm polyethylene, heat-sealed on a diagonal so only the middle of the stems sits in water — leaves kept up to three weeks without spots.

The black spots that make people throw out a bag of kkaennip are not spoilage from age. They are cold damage from the shelf it was stored on.

Two tables of nutrition that do not agree

The United States has no figures for this leaf at all. Searching USDA FoodData Central for “perilla” returns zero entries under Foundation Foods, SR Legacy and Survey — only 41 branded products, mostly oils and seasonings. There is no FDC ID to cite for a raw perilla leaf.

Korea’s own table carries it, and usefully splits it by preparation. From the 7th revision of the Korean Food Composition Table, per 100 g:

	Raw	Blanched	Steamed
Energy (kcal)	35	20	26
Potassium (mg)	389	181	390
Vitamin C (mg)	12	7	6
β-carotene (µg)	9,145	6,891	7,758

Blanching halves the potassium; steaming leaves it untouched. That is a measured difference between two cooking methods — for minerals, which nobody asks about, while the same comparison for aroma does not exist.

One warning about the vitamin C figure. A Korean government food portal publishes 50.6 mg for the same raw leaf, more than four times the 12 mg in the composition table, and gives no source for its numbers. “Perilla leaf is rich in vitamin C” is a sentence resting on an unresolved disagreement.

Perilla oil is not concentrated perilla leaf

Deulgireum comes from the seeds of the same plant — the leaves are picked from June to September and the seed harvested after (RDA, 2020). But the smell is unrelated. Roasting perilla seed at 150–210 °C (300–410 °F) produces an aroma dominated by pyrazines, with 2,5-dimethylpyrazine and furfural as the main individual volatiles (Park et al., 2011). Those are Maillard products, made by the roasting. Perilla ketone, the leaf’s compound, is not what you are smelling in the oil.

What I could not verify

• Any measurement of cooked perilla leaf aroma. Described above. This is the central gap.
• Any comparison, sensory or analytical, of raw against cooked leaf.
• How fast the aroma fades after harvest. A 1986 study measured shelf life and ascorbic acid but not volatiles. A 2017 study states it measured perillaketone during storage, but the results, the storage temperature and the duration are all behind a paywall I could not get through.
• A national production figure for perilla leaf. The Ministry of Agriculture’s vegetable statistics do not list it as a crop; it falls inside “other vegetables”. Every “Korea grows X hectares” figure in circulation traces back to one county’s press releases.
• Leaf thickness or texture measurements for any Korean cultivar.
• Trichome counts on a Korean culinary cultivar. The figures above come from plants grown from Chinese herb-market seed.
• A Korean government source stating plainly that perilla is unrelated to sesame. English guides say it; I could not source it officially.

One measured seasonal detail did turn up, from provincial cultivation trials: leaf hardness fell to 36.5 kg/cm² by mid-August and then jumped to 63.3 by mid-September (Gyeonggi-do ARES, 2004). September leaves are markedly tougher than August ones. Whether they smell different is, again, not something anyone has measured.