The Roast Ledger: What Fire Spends and Earns in a Cup of Oolong

Tea Science｜Author：祥興茶行｜Published：2026-07-09

Roasting turns a brisk, astringent Tieguanyin into something mellow and sweet-finishing. The chemistry behind that change reads like a balance sheet — catechins spent, theanine lost, and Maillard products earned.

Take one batch of Tieguanyin and split it in two. The unroasted, fresh-style half drinks bright and brisk, with a noticeable astringent grip. Roast the other half and it turns round and mellow — the grip fades, and a returning sweetness (hui gan) settles into the finish. So what does roasting actually do to the tea? Think of it as a chemical balance sheet. Heat comes in, and the compounds in the leaf are either spent (they decrease) or earned (they form anew); whatever remains on the books after all that give and take is what you taste. In this article we open that ledger and tally exactly what roasting subtracts — and what it adds.

Meet the Star of the Ledger: Catechins, the Compounds That Make Tea Astringent

Catechins are the main polyphenols in tea leaf. The literature generally puts them at somewhere between ten and thirty percent of the leaf's dry weight, though the figure swings widely with cultivar, season, and leaf position [7]. Think of them as the tea plant's natural sunscreen — the plant makes them to fend off ultraviolet light and insects, and what we taste in the cup is the flavor of that defense.

The catechin family splits into two camps by structure. One camp carries a "galloyl group" — picture a strip of Velcro bolted onto the molecule, ready to grab proteins. These are the ester-type catechins (EGCG, ECG, GCG, and CG are the representatives); the other camp lacks the Velcro and is called non-ester type (EC, EGC, and their kin) [6]. The ester types account for roughly 60 to 75 percent of total catechins, and they are the main drivers of bitterness and astringent grip in the cup [6].

How much does that Velcro matter? Sensory-threshold research has put numbers on it: EGCG registers as astringent at just 0.086 mg/mL, while EGC — identical but for the missing galloyl group — needs 0.16 mg/mL before the tongue picks it up. In other words, the galloylated form makes you pucker far more readily [1]. The same work goes further: EGCG can produce astringency through both a tactile pathway and a taste pathway, while EGC has only the taste route [1].

A caveat: those thresholds and the two-pathway model come from molecular docking and in-vitro sensory experiments. They describe mechanistic tendencies — not an exact count of bonds forming on your actual tongue.

How Astringency Happens on the Tongue

Astringency is not a taste; it is closer to friction. Our saliva carries a family of proline-rich proteins — about seventy percent of total salivary protein [2] — whose day job is keeping the mouth lubricated. Catechins, above all the Velcro-bearing ester types, grab these proteins and bind them into clumps. As the clumps pile up and lubrication fails, that dried-out, drawing sensation in the mouth is what we call astringency [2]. Which is why simply "moving the catechins" is enough to change how smooth a tea feels.

〔圖〕Diagram of how astringency arises: ester-type catechins bind salivary proteins into clumps, lubrication drops, and the mouth feels astringent

The takeaway: astringency can be broken into three layers — abundant ester-type catechins, a strong galloyl group, and salivary proteins being precipitated out. Roasting softens a tea by working on the first layer: the amount and the form of the catechins.

The Debit Column: What Roasting Spends

One study of Tieguanyin roasting kept beautifully clean books. A single batch of tea was split into three versions — unroasted fresh-style, medium roast (107°C for about 210 minutes), and dark roast (a further 115°C for about 150 minutes) — with each component quantified line by line [3]. Start with the two biggest debits.

− The debit column (what decreased)

Debit one: catechins degrade under heat. Total catechins fell from 98.09 mg/g to 90.83 in the medium roast and 80.74 in the dark roast — a drop of about 7.4% and 17.4% respectively [3]. Item by item, every single catechin trended downward: the main drivers of astringency really did shrink.

Debit two: theanine collapses. Theanine is the source of a tea liquor's fresh, savory sweetness, and it gets spent far faster than the catechins — plunging from 18,030 μg/g to just 4,383 μg/g in the dark roast, barely a fifth of the original [3].

Convert both debits into retention rates and set them side by side, and the gap leaps off the page — theanine is lost at practically a free-fall pace.

The takeaway: the main source of bitter astringency (catechins) and the source of freshness (theanine) are both being spent — but at very different speeds. That unequal spending is the setup for the flavor change to come.

The Credit Column: What Roasting Earns

Roasting does not only spend; it also books income. In the same study, the dark-roasted tea recorded several credits [3].

Credit one: products of Maillard and caramelization chemistry. In the dark roast, soluble polysaccharides rose by about twenty percent and gallic acid by about thirty percent, while the Maillard products 5-HMF and DDMP surged to 7.7 and 2.6 times their unroasted levels [3].

Credit two: catechins in disguise — epimerization. Part of what was "spent" did not vanish at all; it was rearranged into a different form.

Credit three: catechins reborn as entirely new flavor molecules. Catechins take part directly in Maillard and browning chemistry, generating new flavor compounds.

Start with credit two, the disguise. Epimerization means a molecule flips one of its "hands" from left to right — same atoms, different three-dimensional arrangement, and with that, a new name. In research on roasted tea, heat converts part of the "epi-type" catechins into "non-epi" forms: some EGCG and ECG turn into GCG and CG [6]. Another Tieguanyin roasting study observed the same direction of travel — as roasting time stretched on, non-epi catechins and gallic acid climbed while epi-type catechins declined [5]. So roasting does not burn the catechins away. A portion degrades, and a portion changes costume and gets booked on the income side.

Why can high heat pull this off? Kinetic studies of heating in water identified two key temperature inflection points, 44°C and 98°C; at the high-temperature end, the epi and non-epi forms settle into a clearly reversible interconversion equilibrium [4].

Know the limits: that kinetic work heated catechins in aqueous solution, while tea roasting is dry heat under very different moisture conditions. It supports the direction and mechanism — high heat drives the interconversion — but its precise conversion rates cannot be transplanted onto a roasting basket [4]. Credit three, the catechin-plus-glucose co-heating, is likewise a pure model system; the specific products still await full-text confirmation and are not equivalent to the complete chemistry of real roasting [8].

Settling the Books: How the Cup Actually Changes

Turn through the ledger page by page, and the final settlement fits into one flowchart — the whole story of this article:

```mermaid flowchart LR A["Roasting heat"] --> B["Debit column catechins ↓ · theanine ↓↓"] A --> C["Credit column Maillard products · GCG conversion · soluble sugars"] B --> D["The tally, as tasted astringency down · bitterness up · hui gan up"] C --> D ```

What do all these debits and credits amount to on the tongue? The study's sensory scores gave an honest answer: roasting made the liquor "thicker in body and stronger in returning sweetness — but bitterness rose in step, and freshness fell" [3].

= The settlement (net change in the cup)

Astringency recedes (fewer catechins), while returning sweetness and body rise — but bitterness actually climbs in step, and freshness drops [3]. This is a ledger of gains and losses, not a one-way upgrade.

One popular story deserves a markdown here. Many people credit the "roasting makes tea sweeter" effect to the Maillard reaction — but in this study, 5-HMF and DDMP are in fact bitter intermediates, not sources of sweetness [3]. The more reliable explanation for why roasted tea drinks sweet and mellow is relative perception: the sources of bitter astringency (catechins) decreased while body increased, and the mellowness shows through. That is correlation — not the Maillard reaction directly minting sugary sweetness.

What It Means for Your Cup

References

  1. Zhou, Y. et al. (2025). Effects of galloyl group on the astringency perception of epigallocatechin gallate and epigallocatechin. _Current Research in Food Science_. https://pmc.ncbi.nlm.nih.gov/articles/PMC12226114/
  2. Effects of theacrine on the astringency of EGCG（含唾液富脯胺酸蛋白澀味機制）(2024). _Food Chemistry: X_. https://pmc.ncbi.nlm.nih.gov/articles/PMC11137526/
  3. Gao, S. et al. (2022). Effects of the Baking Process on the Chemical Composition, Sensory Quality, and Bioactivity of Tieguanyin Oolong Tea. _Frontiers in Nutrition_. DOI: 10.3389/fnut.2022.881865. https://www.frontiersin.org/journals/nutrition/articles/10.3389/fnut.2022.881865/full
  4. Degradation and epimerization kinetics of EGCG in aqueous system (2008) 及 Degradation of green tea catechins. _Journal of Agricultural and Food Chemistry_. DOI: 10.1021/jf0730338；DOI: 10.1021/jf000877h（水相加熱模型，佐證 44°C／98°C 溫度轉折與差向異構化趨勢）
  5. Sensory and chemical characteristics of Tieguanyin oolong tea after roasting. _Food Chemistry: X_. https://www.sciencedirect.com/science/article/pii/S2590157521000663（僅摘要，引方向性結論）
  6. 農研機構（NARO）食品研究部門，茶葉兒茶素定量分析與熱異構化資料（加熱使表型兒茶素部分異構為非表型；酯型/非酯型分類）。https://www.naro.go.jp/laboratory/nfri/introduction/l583.pdf
  7. 農業部茶及飲料作物改良場，《臺灣茶業研究彙報》及一般文獻通則：兒茶素約佔茶葉乾重 10–30%。https://www.tbrs.gov.tw
  8. Model Studies on the Reaction Products Formed at Roasting Temperatures from either Catechin or Tea Powder in the Presence of Glucose (2021). _Journal of Agricultural and Food Chemistry_. DOI: 10.1021/acs.jafc.1c03771（僅摘要，模型體系）

Teas mentioned