Same Leaf, Split Into Green and Black, Kept Almost the Same Caffeine
Catechins collapsed by more than ninety percent between the two halves. Caffeine barely moved. Oxidation, the reaction that defines oolong, is not what decides how much of it ends up in the leaf.
Take one batch of freshly picked leaf. Split it in two. Process one half into green tea and the other into black, the two ends of the oxidation spectrum with oolong sitting between them, and sample both at every stage. Astill and colleagues ran exactly this test in 2001. Catechins in the black-tea half collapsed from 21.34 percent of dry weight to 1.79 percent, a loss of more than ninety percent. Astill names the actor directly: polyphenol oxidase, the leaf's own enzyme, meeting catechins it had been kept apart from and building them into the pigments that darken and thicken a black cup, theaflavins first, then thearubigins. That reaction is the whole of what oxidation is.
Caffeine did not follow the catechins down
It started at 3.18 percent, rose to about 3.6 to 3.8 percent during withering and rolling on both sides of the split, and held there through full oxidation on the black half and through none at all on the green half. Astill's own text is plain about the result: the green half finished at 3.77 percent caffeine, the black half at 3.60, marginally higher on the unoxidized side, not the oxidized one. Whatever oxidation percentage a maker chooses for an oolong, this same-leaf comparison says it is not what decides the caffeine number.
Wuyi rock tea, tested by actual roast level, confirms it
A skeptic could still ask whether oolong specifically behaves differently, since it sits between Astill's two endpoints rather than at either one. Two more recent studies test that directly. Liu and colleagues measured caffeine across four cultivars of Wuyi rock tea2 (yancha, the heavily oxidized, heavily roasted end of the category) at light, medium, and heavy roast: caffeine ran 3 to 16 percent higher at medium roast than at the light and heavy ends, and it did not move in one direction. It peaked in the middle, not at either extreme.
A separate 2026 study tracking nine Taiwanese teas through four fermentation stages reported the same pattern in the plainest terms available: caffeine changes stayed under 20 percent and showed no association with fermentation intensity at all, while the same leaves lost as much as 92 percent of their catechins. A study on CTC black tea processing in North East India found no significant caffeine change during fermentation either, with what small increase there was traced to withering, before oxidation had even begun. Three independent research groups, three different tea categories, one consistent finding: the oxidation step itself runs close to caffeine-neutral.
A Tieguanyin study complicates the tidy version
The Order does not round off the part that argues with the rest. A 2022 study out of Anxi, tracking caffeine at every stage of Tieguanyin manufacture from fresh leaf to finished tea, found a real total decline: 47.7 percent for caffeine, 68.5 percent for tea polyphenols, over the whole process. That is a bigger drop than anything the studies above report, and it lands on the exact style of oolong most likely to be read against this piece.
The decline was not concentrated at one clean step, either. The paper's own stage-by-stage numbers put real caffeine loss across tossing and cooling, the alternating zuoqing cycle that opens the leaf's margins to oxidation, then again through rolling afterward; kill-green itself, the heat that stops oxidation, accounted for only a small further piece of it. So this study does not hand the tidy alternative story (oxidation is innocent, some later heat step is the real cause) either. It says something closer to: for this cultivar, processed this way, real caffeine is lost, and it is lost gradually across most of the sequence, oxidizing steps included, not concentrated where a neat theory would want it. That sits in real tension with Astill's same-leaf split and the Wuyi and Taiwanese studies above, which found caffeine essentially flat through the equivalent stages. The honest reading is that this may be a cultivar-specific or method-specific result rather than a universal one, and the Order has not found a second study that settles which.
The cultivar decides more than the processing does
If oxidation is not the reliable lever, something else sets the number, and the plant itself is the strongest candidate. A 2022 study comparing Chinese-variety cultivars against a low-caffeine Assam-type breeding line found caffeine at 3.58 to 4.15 percent of dry weight in the Chinese material and as low as 0.05 to 0.55 percent in the Assam line, traced to near-absent expression of TCS1, the gene coding for tea caffeine synthase, itself suppressed by regulatory proteins specific to that breeding line. That is a wider caffeine range than any processing study above produces, set by the cultivar under the picker's shears, not by anything the maker does to the leaf afterward.
A study across six Kenyan cultivars found the same shape at a smaller scale, with processing removed from the comparison entirely: researchers measured caffeine straight from the fresh, unprocessed shoot, cultivar by cultivar, and it still varied by variety, and again by season, running highest in the hot, dry months. A separate survey of thirty tea infusions put oolong's own dry-weight caffeine at 12.36 to 31.66 mg per gram of dry leaf, the widest spread of any of the six categories tested: green, black, white, yellow, and dark included. Camellia sinensis is not one uniform caffeine source dressed up differently by processing. It is many genotypes, each with its own baseline, and oxidation and roast sit on top of that baseline rather than setting it.
The pluck matters too
Layered on top of cultivar is which part of the plant went into the basket. A 2012 study of tea grown in Hawaii reported caffeine and l-theanine both declining as leaf age increased, from bud through the first leaf to the second, while catechins ran the other way, rising with age. Caffeine's usual job in the plant is defensive, deterring insects from the newest, most vulnerable growth, so a bud-heavy pluck should read higher in caffeine than a coarser one from the same cultivar and the same oxidation level. The exact size of that gradient is not yet pinned down as precisely as the cultivar figures above, so the direction, not a specific percentage, is the safe claim.
Oxidation percentage is the number every oolong label leads with, and it is the variable least likely to explain a real difference in caffeine between two teas. Cultivar explains far more of it, the pluck adds another layer, and at least one study on Tieguanyin says the manufacturing sequence itself is not entirely blameless, just not in the single clean spot a tidy theory would prefer. What lands in the cup after all of that is a separate question again, governed by water temperature and steep time; the brewing chemistry behind why a green-leaning oolong and a dark roast want different water already covers that ground. The leaf decides how much caffeine it holds. The kettle decides how much of it you get.