Putting “not astringent” and “toasty” into the language of compounds

Chapter 8 · 4 of 4 sections readable now

Illustration of hojicha with curved lines representing its aroma and an open notebook containing sketches of tea leaves

What makes tea astringent falls sharply in the roaster. What remains has changed as well: more of it is now in forms whose names have lost the “epi” at the front.

Why hojicha is less astringent: what happens to the catechins

Roasting does more than lower the source of astringency: part of it turns into another form. This piece follows how that change works, and where the taste of the new forms is still untested.

An illustration of two white teacups on a wooden dinner table, one of amber hojicha and one of pale green sencha, with small dishes of brown and green tea leaves in front.

After dinner the cup comes up to the lips. One mouthful goes down, and the tongue is left clean. Sencha poured into the same cup made a child pull a face. The packet says no more than “mellow” and “soft astringency”.

What follows traces how the source of astringency rearranges itself in the roast, and how much is known about the taste on the far side.

What settles at the bottom of the cup

Set the cup down and a pale brown gathers at the bottom. Roasting lowers the source of astringency, and changes its line-up as well.

What packets and articles call tannin is, in tea, a group of compounds called catechins. An earlier chapter followed their total falling to about a fifth in the roast. The same chapter noted that some of them rise along the way.

The falling total is the familiar part. The sources go as far as showing that the rising ones have a definite face. That face comes next.

The “epi” falls from the front

Pinch a roasted leaf and it crumbles with a dry crackle and a toasty smell. The compounds that rise are the forms without “epi” at the front of their names.

A prefectural research station roasted tea light, medium and dark, then measured the catechins compound by compound. The most abundant one fell sharply, while others climbed steeply. Before roasting almost all are epi forms; by the light roast, more than a third have switched.

The falling total gets passed around, while this change of address rarely does. The amounts sit in the table below, and the sources go as far as the switch itself. The switch, as it happens, has a name.

Figure 8-1-1The eight catechins and the share of each form, by depth of roast (Kochi, first-flush tea; mg per 100 mL of infusion)

Compound (first-flush tea)	Aracha	Light roast	Medium roast	Dark roast
Epicatechin (EC)	7.43	5.17	3.41	1.98
Catechin (C)*	0.17	0.28	0.20	0.09
Epigallocatechin (EGC)	2.09	0.50	0.96	0.72
Gallocatechin (GC)*	0.00	0.62	0.39	0.00
Epicatechin gallate (ECg)	3.50	1.88	1.25	0.53
Catechin gallate (CG)*	0.86	1.89	1.70	1.11
Epigallocatechin gallate (EGCg)	20.38	6.29	3.15	1.18
Gallocatechin gallate (GCg)*	0.46	5.23	3.53	1.38
Total	34.89	21.86	14.59	6.98
Non-epi share [calculated]	4.3%	36.7%	39.9%	36.9%
Gallate-type share [calculated]	72.2%	69.9%	66.0%	60.1%

From Table 1 of “Roasting Level and the Quality of Hojicha”, Tea Research Station, Kochi Prefectural Agricultural Technology Center (data year 2013). Sample = first-flush aracha and the light, medium and dark roasts made from it by roasting twice in an automatic continuous hojicha roaster. Unit = mg per 100 mL of tea liquor; brewing = 1.9 g of leaf in 200 mL of boiling water for 5 minutes. The exhaust temperature at the second roasting was 100°C for all three roasts; only the drum's rotation speed was varied (chapter 6). The temperature of the leaf itself was not recorded. * = the four non-epi compounds that rise with roasting. The two share rows are calculated and are not printed in the source: non-epi = C + GC + CG + GCg, gallate-type = ECg + CG + EGCg + GCg. The epi-type share (EC + EGC + ECg + EGCg) is 100% minus the non-epi share. In karibancha the non-epi share goes from 3.0% to 36.4%, in second-flush tea from 3.4% to 42.4%, moving the same way. The dark-roast second-flush tea totals 8.68 mg, more than the 6.98 mg of the dark-roast first flush. The course of the total already appears in Figure 2-2-1 of chapter 2; what is shown here is the breakdown into eight compounds and the shares.

It happens in kitchen water too

Steam from the kettle clouds the kitchen window. Under heat, a catechin flips at one single point in its shape.

Like a glove turned inside out, each has a fixed partner: the same name with the “epi” taken off. A review of the chemistry shows every pair working the same way, matching the compounds that rose in the station's table. An old report had already written the rearrangement down.

Warming catechins in water is enough to set the flip going. Roasting, though, puts dry leaf over heat, and the sources only go as far as saying the conditions differ. What is left to ask is how the flipped forms taste.

A taste no one has measured yet

A cooling mouthful rolls towards the back of the tongue. How astringent the flipped forms are has not yet been tested.

A review that tasted catechins at tea strength found the strongest tastes among the epi forms. One leaned astringent, the other bitter. Only epi forms were tasted there; the ones that rise in the roast were left out.

No study found here measures the rising forms at the strength of a hojicha infusion. On the taste side, “roasting softens astringency” is short of support, and the sources reach only as far as the switch. Last comes the raw leaf itself.

A dark roast from a later picking

On the shelf stand dark roasts from an early picking and a later one. Change the raw leaf, and the amount moves even at the same depth.

Roasted to the same depth, the second picking kept more catechins than the first. Matching the roast does not match what is inside.

The sources go as far as saying depth alone does not fix the amount. The matching happens further along, and a later part takes up blending. The figures and test conditions are gathered just below.

The numbers in detail

The trial: The trial is “Roasting Level and the Quality of Hojicha” by the Tea Research Station of the Kochi Prefectural Agricultural Technology Center. Aracha from first-flush, karibancha (刈番茶) and second-flush crops was roasted twice in a hojicha roaster into light, medium and dark roasts. The catechins were measured as eight compounds.
Totals and breakdown: The fall of the total to about a fifth appears in chapter 2, Figure 2-2-1. Epigallocatechin gallate, the most abundant, fell sharply, while one compound rose more than elevenfold. The four that rise are the forms with no “epi” at the front of the name. The values by compound are in Figure 8-1-1.
Share by form: The proportions by form are calculated values not printed in the original. In aracha, 95.7% of the catechins are epi-type; by the light roast, non-epi forms hold 36.7%.
The mechanism: The mechanism follows a 2007 review by Sano. Heating inverts the catechin ring at position 2 alone, so epigallocatechin gallate becomes gallocatechin gallate. All four pairs work the same way, matching the four that rose in the Kochi table.
Older report, warm water: A 1968 report by Nakagawa had already described the rearrangement as epimerisation, polymerisation and decomposition. By the same review, standard catechins steeped for 30 minutes in a buffer at 90°C (pH 6.0) isomerise.
Taste review: The taste data come from a 2022 review by Narukawa. Four epi-type catechins were tasted at the concentrations found in tea infusions (300 micromoles per litre and below). The strongest was epicatechin gallate, then epigallocatechin gallate. The first is described as more astringent, the second as more bitter.
Raw-leaf difference: The dark-roast second flush held 8.68 mg of catechins in total, more than the 6.98 mg of the dark-roast first flush. Both are per 100 mL of infusion.

Sources

Updated 17 October 2026This section on its own page

02Members only

Hojicha aroma compounds: which ones are doing the work

No single compound makes hojicha's aroma. This piece follows it through three words, toasty, sweet and green, and through why the compounds at work differ between the steam and the cup.

Open the bag and a dry toastiness drifts out. Pour water into the pot and the steam smells a little sweeter. The leaf is the same, yet the word that comes to mind changes each time.

What follows traces the compounds behind the aroma, from the leaf to the steam to the liquid in the cup.

Toasty, sweet and green

Cupped in both hands, the tea sends steam past the nose. No single compound decides the aroma; three words bundle it instead.

Toasty, sweet, green. The research, too, gathers the smells under these three. Of them, the toastiness and the sweetness are what the fire made.

No source names one compound as the lead, and the most that can be said is that the smells fall into three bundles. Next comes what was measured in the leaf, the steam and the liquid.

Dry leaf and rising steam

Bring the nose to the dry leaf and the toastiness reaches far back. Change where it is measured, and the compounds at the top change too.

One method dilutes an aroma step by step and ranks compounds by how far they can still be smelled. In the leaf itself, a toasty compound and a sweet one shared the top step. The steam, and the liquid with the leaf removed, were measured the same way.

In the steam only one compound reached the top, while the liquid held more there. The sources go as far as saying the line-up shifts with the place, even within one tea. The reason lies with water, steam and the nose.

The remaining four sections follow a faint compound into the steam, explore how fire forms most toasty and sweet aromas, and connect compound names with smells. Figures and conditions close the piece.

The rest of this article is for members.

Membership is available in Japanese. It takes an email address and a password, and it is free.

Open the Japanese membership page

Sources

Updated 28 October 2026This section on its own page

Does hojicha go stale? What falls first

How many days a freshly roasted aroma lasts has not been measured. What is known is the order in which things fade. This piece follows aroma, the source of astringency, borrowed smells and what sellers advise.

An illustration of a lidded tea tin, an open packet of tea, and a tea scoop on a shelf.

From the back of the shelf comes a bag with its mouth held shut by a clip. When it was opened is long forgotten. Bring a nose close and there is still a smell. The thing to compare it with, though, went first.

What follows traces what fades first from stored tea, through a storage trial and what tea sellers publish.

The smell as the clip comes off

Take the clip off and a faint toastiness reaches the nose. The days cannot be counted; what can be told is the order of the fading.

The aroma and the look of the leaf go first, while the source of astringency barely moves. That order can be read from a trial that kept sencha for a long time.

No source measuring the days for hojicha turned up, and the sources go only as far as the order. Next comes the row of bags in that trial.

Sencha bags in a dark cupboard

In a dark cupboard, bags of the same sencha stand in a row. The worse the conditions, the faster the aroma falls.

The trial combined how well the bag kept gas out, whether the air inside was replaced, and temperature, then assessed the tea at intervals. Tea kept in a warm room in a clear bag reached its lowest point within a year. An aluminium bag slowed the fall, and chilled tea held up highest to the end.

The conclusion: chill the tea for a short stay, and for a long one combine a good bag, replaced air and the fridge. The sample was sencha, and the sources speak only to sencha's aroma. After the aroma come the compounds.

Figure 8-3-1Aroma assessment score by storage condition (sencha, out of 20)

Storage condition	6 months	1 year	18 months
Control (−80°C, gas-flushed, aluminium bag)	14	14	14
25°C, air, transparent bag (No. 4)	13	8	9
25°C, air, aluminium bag (No. 1)	13	10	10
5°C, air (four bags)	13–14	12	11–12

The sample is sencha, not hojicha. From Table 2 of Mizukami and Yamaguchi (2013), Tea Research Journal 116. Sample = lightly fired sencha made from Yabukita plucked on 24 April 2009, 50 g each, stored in the dark. The values are the aroma scores in the standard assessment method (each item out of 20 points), a sensory evaluation, not a measurement of compounds. The aroma score before storage was 14. The compounds (catechins, caffeine, amino acids) were measured at six months and one year only; at eighteen months only the sensory assessment was carried out. The transparent bag (No. 4) has the lowest gas barrier of the four bags. The 5°C row gives the range across the four bags.

The source of astringency stays put

Tea from the same bag is brewed, and the cup is peered into. Catechins, caffeine and amino acids barely move in storage.

Even in the worst condition, the warm room, the report found no marked change. The compounds, though, were measured only partway; the final assessment was by tongue and nose alone.

And this is a result for sencha, with no promise that the same figures hold for hojicha. The authors themselves say it cannot be carried straight into practice. That is as far as the sources go. Next comes the aroma mixing with other smells.

A smell that fades, a smell that is borrowed

A bag of tea sits next to a spice jar. An aroma that fades and a smell that is taken up from elsewhere are two different failures.

The assessment vocabulary for Japanese tea lists the two as separate words. Tea held warm or left long in air carries a damp or deteriorated smell. Tea that has taken up other smells carries a transferred or foreign smell.

These words, though, were settled for storing aracha, in the assessment of sencha and gyokuro. No public vocabulary for hojicha's aroma turned up, and that is as far as the sources go. Last comes what the sellers advise.

Not the days, but the place

On the back of a shop's bag, words about where to keep it run in small print. What sellers publish leans towards the place rather than the days.

Long-established tea shops, too, list air, moisture, light and changes in temperature as things to avoid. One shop writes that it roasts just before shipping and seals the tea into small packs. None of them says how many days a freshly roasted aroma lasts.

Some give a best-before date for unopened tea, but that is a product's shelf life, not the speed at which aroma falls. The sources go as far as the order of the fading. Where to keep it is taken up in a later chapter. The figures and test conditions are gathered just below.

The numbers in detail

Storage trial: The storage trial is the 2013 report by Mizukami and Yamaguchi (Tea Research Journal 116). Lightly fired sencha from Yabukita plucked on 24 April 2009 was kept in the dark. Bag barrier properties, gas flushing or not, and 25°C or 5°C were combined, with standard assessment every six months.
Aroma scores: The aroma score is out of 20, and the tea scored 14 before storage. Tea kept at 25°C in air in a transparent bag reached its lowest point within a year. At 5°C it still scored 11 to 12 at eighteen months. The values are in Figure 8-3-1.
Conclusion: The conclusion is that six months calls for 5°C refrigeration, and a year calls for a high-barrier bag, gas flushing and 5°C.
Compound tests: The compounds (catechins, caffeine and amino acids) were measured at six months and one year. At eighteen months only the sensory assessment was done. No marked change was found even in the poor environment of 25°C in air.
Assessment terms: The assessment vocabulary is from “Current status and prospects of Japanese-tea quality evaluation” by the NARO Institute of Vegetable and Tea Science. It was set out for the storage of aracha, as terms for assessing ordinary sencha and gyokuro.
Seller advice: Maruhachi Seichajo names air, moisture, light and changes in temperature. Ippodo Tea advises keeping tea clear of sunlight, heat, moisture and strong smells. Bashoen Chaho says it roasts just before shipping and seals the tea into individual packs.
Best-before date: Maruhachi Seichajo gives a best-before date of three months from manufacture, unopened (chapter 7). Where to keep tea is taken up in chapter 12.

Sources

Updated 17 October 2026This section on its own page

04Members only

What the research covers, and what it leaves out

“Roasting destroys the astringent compounds”: how much of that sentence has been measured? This piece takes astringency, sweetness and aroma in turn, and separates where research reaches from where it does not.

An illustration of a wooden desk with three plain, unlabelled booklets stacked on it, beside a cup of hojicha and an empty white saucer.

Books and articles sometimes put it like this. Hojicha is not astringent because the heat of roasting destroys the astringent compounds. One sentence settles it, and the reader nods along.

What follows stands in front of that sentence with the figures gathered so far, and traces what has actually been measured.

Past the sentence that earned a nod

The book is closed partway, and a hand goes out to the cup. What has been measured goes as far as the compounds changing places.

Whether the new forms are astringent at hojicha strength has not been tested. The word “destroys” stands a step beyond what was measured.

The sources go as far as the swap. Next come the compounds that were actually tasted.

The epi forms on the tongue

A mouthful is held, waiting for anything left at the sides of the tongue. At tea strength, only the forms with “epi” have been tasted.

For one compound on the rising side, there is a study that rated its bitterness and astringency over time. That study, though, was not about tea. No study found here tests the rising forms at the strength and under the conditions of a hojicha infusion.

On top of that, a review notes that much about how astringency is sensed remains unclear. The sensation is thought to mix bitterness with a physical feel in the mouth. The sources go as far as saying the sense itself is still finding its footing. Next comes sweetness.

The four sections that remain untie “sweet” into separate measurements, trace where the theanine went, and weigh the sources while leaving the unfilled gaps open. The figures close the piece.

The rest of this article is for members.

Membership is available in Japanese. It takes an email address and a password, and it is free.

Open the Japanese membership page

Sources

Updated 17 October 2026This section on its own page

Column

Writing an aroma down in fifteen words

The dilution factors set out in “Hojicha aroma compounds: which ones are doing the work” are made like this. An aroma extract is taken from the leaf and diluted tenfold at a time. A machine separates the compounds, and one branch of its outlet is placed at the height of a person's nose. Someone sits there, and the last dilution at which they can still say they smell something becomes that compound's figure. Ten, a hundred, a thousand, ten thousand. What the numbers really record is the number of times a diluted liquid was smelled by a person.

Beside the numbers is a column of words: what the person smelling called the smell. Count the words attached to the fifty compounds and there are fifteen. Four of them — metallic, fecal, fatty and off-odour — belong to the side that names faults. The paper notes of this vocabulary that it was presented with reference to a great many papers and to the descriptive terms used for food aromas.

The words move with concentration. Furfuryl mercaptan, identified in tea for the first time in this study, is described as smelling of garlic chives when concentrated and of coffee when dilute, and the word it was given in this analysis was the latter. The same molecule answers to a different name depending on how much of it there is.

The dilution steps differ from study to study. A dilution factor therefore means nothing placed beside a figure from another paper. What can be placed side by side is the order within one study, and these fifteen words.

Sources: Mizukami Y. (2012), “Identification of odour-active compounds in hojicha leaf by high-vacuum distillation with a SAFE apparatus and aroma extract dilution analysis”, Tea Research Journal 113:55-62 / https://www.jstage.jst.go.jp/article/cha/2012/113/2012_113_55/_pdf / verified 2026-09-24

Back to the contents