Japanese milk bread, shokupan, is soft, tall, and famously slow to go stale. The technique credited with all three is a cooked flour paste stirred into the dough: yudane in Japanese practice, tangzhong in the version that spread through English-language baking. Boil a small portion of flour with water, cool it, add it to the rest. The explanation offered is always the same, that gelatinised starch holds more water and slows staling.
The starch physics behind that explanation is genuinely well characterised. The method itself is one of the least documented widely-practised techniques in baking. Separating the two is the useful thing to do here.
Where wheat starch actually gelatinises
Gelatinisation is measured by differential scanning calorimetry, which reports three temperatures: onset, peak, and conclusion.
For wheat starch, a 2025 study in Foods measuring a control starch dispersion, roughly 10 milligrams of starch with 30 microlitres of water, heated at 5 degrees Celsius per minute, reported an onset of 55.4 degrees with an enthalpy of 7.16 joules per gram. A 2024 study in Frontiers in Nutrition, heating at 10 degrees per minute after equilibrating samples at minus 4 degrees for 24 hours, reported onsets of 58.4 to 60.7 degrees, peaks of 61.6 to 63.3, and conclusions of 63.7 to 65.7 across three growing treatments. A 2025 paper in Current Research in Food Science working with dilute systems put the wheat peak at 61.5 degrees.
Why those numbers disagree
The peak values from two independent studies overlap closely, around 61 to 63 degrees, and the conclusion values overlap as well. The onsets do not: 55.4 against 58.4 to 60.7 is a real gap.
The likely explanation is in the method rather than in the starch. Heating rate differed by a factor of two, water to starch ratios differed, and one study pre-equilibrated below freezing. Enthalpy differed even more dramatically, 1.2 to 1.5 joules per gram in one study against 7.16 in the other, which is too large a spread to be a property of wheat and is instead a property of the measurement system. This is the ordinary situation in cereal science, and it is why a single quoted gelatinisation temperature for wheat starch should be treated with suspicion.
Gelatinisation finishes later than the peak suggests
A more consequential finding, for anyone cooking a paste on the stove, comes from a 2018 study in Scientific Reports. Working across water to starch ratios from 2 to 4, it reported wheat conclusion temperatures of 65.6 to 67.5 degrees, and then made a distinct point: at the conclusion temperature, wheat and maize starches still retain structural order. Rice starch, at ratios of 2 to 1 or above, is fully gelatinised by then, but wheat is not.
The authors proposed a further parameter, an end temperature at which the calorimetry trace returns to baseline, and reported it for wheat at 71.0 to 73.6 degrees. Their conclusion was that gelatinisation of wheat and maize starches is more complete at that end temperature than at the conclusion temperature.
For a cook, the implication is direct. Heating a flour and water slurry to around 65 degrees, which many yudane instructions specify, lands in the region where the DSC peak has passed but ordered structure remains. Whether that matters for the finished bread is an open question, but it is not the same thing as full gelatinisation.
The same body of work also clarifies a division of labour between the two starch fractions. Amylose content, the 2025 dilute-system study reports, appears not to be important for the phase transitions themselves, but governs gel formation on cooling. So the swelling happens largely through amylopectin, and the setting of the cooled paste is amylose's contribution.
Staling is retrogradation, and the fridge makes it worse
The second half of the yudane claim concerns keeping quality, and here the mechanism is well established.
A 2021 study in Foods on bread storage conditions sets out the temperature window: in starch gels at bread-like concentrations, retrogradation occurs above the glass transition temperature, around minus 5 degrees Celsius, and below the melting temperature, around 60. Recrystallisation involves two processes, nucleation and propagation, whose rates peak at opposite ends of that window, near the glass transition and near the melting temperature respectively. Citing earlier work by Slade and Levine, the study states that the combined retrogradation rate is higher at refrigerator temperatures, so storing bread in a fridge accelerates firming.
That is one of the few pieces of bread science that contradicts intuition and is worth acting on. Cold does not preserve crumb softness; it speeds up the process that destroys it.
Retrogradation is also measurable in the same way gelatinisation is. The 2025 Foods study reported, for its wheat starch control, a retrogradation melting onset at 37.1 degrees with an enthalpy of 2.41 joules per gram, and found that adding proteins raised that onset to between 41.5 and 43.7 degrees depending on the protein.
What the published work on yudane actually says
Almost nothing that can be read.
Two peer-reviewed papers address the technique directly. One, in the Journal of Texture Studies in 2017, is titled for the effect of a water roux starter, given as tangzhong, on the texture and consumer acceptance of rice pan bread. The other, in the Journal of Cereal Science in 2023, concerns yudane bread produced using amorphous wheat flour processed in a shear and heat milling machine. Both are behind paywalls, and in the first case the abstract itself is withheld by the publisher.
A third paper, in a 2024 issue of a Korean culinary science journal, studies tangzhong pan bread but varies the addition of an insect powder rather than the tangzhong itself. Its useful contribution is showing which measurements the field considers relevant: pH, volume, weight, specific volume, baking loss, texture profile analysis, colour, water content, and sensory evaluation.
So the honest position is this. The figures that circulate for how much extra water a tangzhong lets a dough carry, how much specific volume improves, or how much slower the crumb firms are not, as far as publicly readable literature goes, traceable to a measurement. They may well be correct. They are not currently checkable.
An adjacent study pointing the same way
The closest readable evidence comes from the same research group behind the 2023 yudane paper. A 2025 study in the Journal of Applied Glycoscience, by authors at Yamagata University, develops a rice bread method by controlling gelatinisation temperature with modified starch, focusing on hydroxypropylated potato starch.
Its stated rationale is the interesting part: that modification not only suppresses retrogradation but also lowers the gelatinisation temperature. That is exactly the pair of effects a yudane is supposed to deliver, pursued by a different route. In their trials, batters with the modified starch progressively suppressed the large central cavity that formed in controls, with the most substituted sample performing best. The bread was fermented at 40 degrees for 40 minutes, baked at 200 degrees for 40 minutes, and cooled at room temperature for 24 hours before assessment.
This does not validate the yudane method. It does show that lowering gelatinisation temperature and suppressing retrogradation are treated as a coherent, testable pair of objectives by researchers working in this area.
Reading the method honestly
If you bake shokupan with a yudane and the crumb is softer and lasts longer, that observation stands on its own. The proposed mechanism is plausible and rests on well-measured starch behaviour. What does not exist, in readable form, is the quantitative bridge between the two.
Meanwhile the one storage finding that is well documented costs nothing to apply: keep the loaf out of the fridge.
