Chai and concrete both depend on the order you add things because each contains a fast reaction that happens to whatever is present first and cannot be undone. In chai, ginger boiled in water before the milk goes in loses its milk-clotting enzyme, while ginger added to warming milk can curdle it in about 90 seconds. In concrete, water added late weakens the cured block by Abrams' law, roughly 27% for every extra 0.1 of water-cement ratio, and a plasticiser added 10 to 15 minutes after mixing works better than one added at once.
- Path dependence: same final recipe, different history, different product, in both chai and concrete.
- 40–60°C is where ginger protease cuts milk protein fastest; it clots milk best at 70°C.
- About 90 s in that window is enough for ginger juice to start curdling milk.
- Abrams' law (1918): concrete strength falls exponentially with the water-cement ratio.
- +0.1 w/c costs about 27% of compressive strength by Abrams' own equation.
- 10–15 min delay is the best moment to add a naphthalene plasticiser, per a 1999 study.
Every chai-wala has a sequence and defends it. Water, ginger and spices first, leaf next, milk last, sugar whenever. A site engineer has one too, written into the mix design: aggregates, cement, most of the water, admixture later, and never a bucket of water at the end to make it pour easier.
Both are right, and for the same reason. This post solves the two kitchens, the stove and the mixer drum, with real numbers, and then shows exactly where the analogy stops.
On 12 January 1946, George Orwell published "A Nice Cup of Tea" in the London Evening Standard, listing eleven rules he called golden. Rule ten took a side in Britain's oldest kitchen argument: pour the tea first, then the milk, because that way you can judge exactly how much milk goes in. He insisted Indian or Ceylon tea was the only kind worth the name. In June 2003, for Orwell's centenary, the Royal Society of Chemistry released a counter-recipe written with Loughborough University chemical engineer Andrew Stapley. Milk first, it said, because milk poured into hot tea comes apart into drops that briefly meet very hot liquid, and milk proteins degrade above about 75°C. Orwell argued about control; the chemists argued about history. Both agreed that order matters.
The Orwell Foundation: A Nice Cup of Tea (1946)Why does the order you add things matter at all?
Order matters whenever a fast, irreversible step can happen to only some of the ingredients: whatever arrives first takes part, and what arrives later finds the step already done. Physicists call this path dependence. A mixture's final composition does not tell you its properties; its history does.
Three kinds of irreversible step do the work in these two recipes:
- Enzyme destruction. Heat unfolds an enzyme for good. Boil ginger first and its protease is dead before it ever meets milk.
- Binding. Molecules that stick to the first surface they meet are not available for the next. Milk casein grabs tea catechins; cement's earliest minerals grab plasticiser.
- Dilution of a strength-setting ratio. In concrete the water-cement ratio decides the pore space in the hardened paste. Water added after the fact still counts.

Why does milk curdle if ginger goes in at the wrong time?
Ginger contains a protease, a protein-cutting enzyme, that snips the κ-casein holding milk's casein micelles apart, exactly as cheese rennet does. It works best between 40 and 70°C. Milk heated slowly with ginger spends long enough in that band to start splitting; ginger boiled in water first is denatured before the milk arrives.
The China Agricultural University team extracted proteases from fresh ginger rhizome and purified them by ion-exchange chromatography. They found three forms, all about 31 kDa, with optimal protein-cutting activity between 40 and 60°C and maximum milk-clotting activity at 70°C. The enzymes cut αs1-, β- and κ-casein, with the most specific cuts on κ-casein at two preferred bonds. The authors note ginger is already used as a milk coagulant in the Chinese dessert ginger milk curd. The limit: purified enzyme in buffered milk, not grated ginger in a pan.
Read the paperReconstituted milk (12% milk powder, 10% sugar, pH 6.29) was heated to 95°C for 5 minutes, cooled to 61°C and mixed with 5% ginger juice. No curd had formed at 60 seconds; it began at 90 seconds, was largely formed by 5 minutes and complete by 10. Protease activity peaked at 13.11 U within the first minute and then fell, which the authors attribute to heat inactivation. Final hardness was about 305 gf. The limit: one recipe at one temperature, so it does not map how fast the enzyme dies at 100°C.
Read the paper (free)Worked example: how long does milk sit in the danger band?
Take a two-cup chai: 150 ml water plus 150 ml milk, heated together with crushed ginger from 25°C. Assume the stove raises the mix about 20°C a minute, which is my estimate for a domestic gas burner; check it with the calculator below.
- Energy the pan must absorb. With \( m \approx 0.3 \) kg and \( c \approx 4.0 \) kJ/kg·K (close to water),
\[ Q = m c \Delta T = 0.3 \times 4000 \times (100 - 25) = 90{,}000\ \text{J} \]
- Heating rate. At 20°C a minute, 75 degrees take 3.75 minutes, so the burner is delivering
\[ P = \frac{90{,}000}{225\ \text{s}} = 400\ \text{W into the liquid} \]
- Time in the enzyme's window. The protease is most active from 40 to 70°C:
\[ t_{window} = \frac{70 - 40}{20\ ^\circ\text{C/min}} = 1.5\ \text{min} = 90\ \text{s} \]
- Compare with the curd clock. In the 2025 experiment curd started forming at 90 seconds. My heated pan spends the same 90 seconds in the active band. That is a coin-flip, which matches the kitchen experience: sometimes it splits, sometimes it does not.
Now reverse the order. Boil the ginger in the 150 ml of water for two minutes first. The enzyme sits at 100°C, far above the 70°C clotting peak, and the 2025 data show its activity already falling at 61°C. When cold milk is poured in, the mix drops to about 62°C and climbs back to a boil, but there is little active enzyme left to cut anything.
Does water first also make chai stronger?
Yes, in the sense that the leaf gives up its caffeine and catechins quickly into plain hot water, before milk proteins are around to bind them. Black tea releases most of its caffeine and catechins within 2 minutes and its total polyphenols by 6 to 8 minutes.
Six independent brews of 5 g Sri Lankan CTC black tea in 500 ml of just-boiled water were sampled at 0, 1, 2, 4, 6, 8, 10, 12, 14 and 20 minutes while the water cooled from about 100 to 75°C. Caffeine, gallic acid and catechins came out mostly within 2 minutes; total phenolics and flavonoids peaked at 6 to 8 minutes. Caffeine capacity was about 42 mg per gram of leaf. The authors recommend 2 to 8 minutes. The limit: one tea, plain water, no milk and no rolling boil.
Read the paper (free)What milk does to those catechins was shown in a small human trial. In 2007, 16 women drank black tea, black tea with 10% skimmed milk, or water. Tea improved blood-vessel dilation; tea with milk did not, and lab work traced the difference to caseins forming complexes with tea catechins. For chai, that binding is partly the point, because it softens the astringency of strong CTC leaf. Order decides how much flavour comes out of the leaf before the casein starts mopping it up.
| Step in chai | Fast irreversible event | Concrete's version |
|---|---|---|
| Ginger into boiling water | Protease denatured above 70°C | Admixture timed to the first minutes of hydration |
| Leaf into plain water | Catechins extracted in about 2 min | Cement wetted evenly before admixture |
| Milk added | Casein binds catechins | Fast early reactions adsorb plasticiser |
| Milk poured into very hot tea | Drops briefly exceed 75°C | Admixture dumped on dry cement |
| Extra water at the end | Weaker, thinner cup | Higher w/c ratio, weaker concrete |
What does Abrams' law say about adding water to concrete?
Abrams' law says the compressive strength of a workable concrete depends on just one proportion, the ratio of water to cement, and falls exponentially as that ratio rises. Duff Abrams of the Lewis Institute in Chicago drew it from about 50,000 tests and published it in 1918.
Abrams reports roughly 50,000 tests, mainly compression tests of 6 by 12 inch cylinders at 28 days, covering mixes from 1:15 to neat cement, many aggregate gradings and consistencies. Plotted against the water-cement ratio, the scattered points fell on one smooth curve, \( S = A/B^x \), where \(x\) is the volume of water per volume of cement. For his 1:4 test mix he gave A = 14,000 psi and B = 8.2. He concluded that aggregate size and cement content matter only through the water they force you to add. The limit: 1918 cements, so absolute strengths are lower than today's; the shape of the curve still holds for workable mixes.
Read the bulletin (free)
Worked example: the bucket of water at the end
- Convert Abrams' ratio to modern units. His \(x\) is by volume, with 94 lb of cement to a cubic foot and water at 62.4 lb per cubic foot, so the usual mass ratio \(w/c\) gives
\[ x = \frac{w}{c} \times \frac{94}{62.4} = 1.506\,\frac{w}{c} \]
- Strength at w/c = 0.45.
\[ S = \frac{14{,}000}{8.2^{\,0.678}} = \frac{14{,}000}{4.16} \approx 3{,}360\ \text{psi} \approx 23.2\ \text{MPa} \]
- Sensitivity. Differentiate the log of Abrams' law:
\[ \frac{dS}{S} = -\ln(8.2)\,dx = -2.104 \times 1.506\, d\!\left(\tfrac{w}{c}\right) \;\Rightarrow\; \frac{S_2}{S_1} = 8.2^{-1.506\,\Delta(w/c)} \]For \( \Delta(w/c) = 0.1 \), \( S_2/S_1 = 0.73 \): every extra 0.1 of water-cement ratio costs about 27% of the strength.
- The site case. A cubic metre with 350 kg of cement at w/c 0.45 holds 157.5 litres of water (my assumed mix). Add 10 litres to help it pour: w/c becomes 0.479 and
\[ \frac{S_2}{S_1} = 8.2^{-1.506 \times 0.0286} = 0.91 \]That bucket cost about 9% of the design strength, permanently.
See the data as a table
| Abrams 1:4 mix (S = 14,000/8.2^x psi) (MPa) | |
|---|---|
| 0.35 | 31.8 |
| 0.40 | 27.2 |
| 0.45 | 23.2 |
| 0.50 | 19.8 |
| 0.55 | 16.9 |
| 0.60 | 14.4 |
| 0.65 | 12.3 |
| 0.70 | 10.5 |
| 0.80 | 7.6 |
Source: Abrams, Design of Concrete Mixtures, Lewis Institute Bulletin 1 (1918); converted to MPa and mass ratio · Calculated from Abrams' published constants for his 1918 test mix. Modern cements are stronger; the shape is the point.
Why is the water so costly? Cement needs only about 0.30 to 0.35 of its weight in water to hydrate fully. Typical concrete uses 0.40 to 0.60 so it can be placed. Every litre beyond what the chemistry consumes stays behind as pores, and porous paste is weaker and lets water and chlorides in.
Why do builders add the plasticiser late?
A plasticiser works by sticking to cement grains and pushing them apart. Added with the first water, much of it is swallowed by the fastest-reacting minerals in the first minutes; added 10 to 15 minutes later, once those early reactions have slowed, more of it stays on the grains where it is needed.
The National Taiwan Normal University team added a naphthalene-sulfonate superplasticizer (SNF) to Type I Portland cement at different times after mixing and measured how much adsorbed and how the concrete flowed. Adsorption fitted a Langmuir isotherm. As addition was delayed, saturated adsorption dropped sharply at first and then slowly, while workability changed in step, with both transitions at about 10 to 15 minutes. They put the best addition time in that window, at the start of cement's dormant period. The limit: one cement and one admixture type; modern polycarboxylate admixtures behave somewhat differently.
Read the abstractThe Langmuir model the authors used describes surface binding:
Here \( \Gamma \) is the amount stuck per gram of cement, \( C \) the concentration left in the water and \( K \) the binding strength. In the first minutes the cement surface is reacting fastest, so an admixture added at minute zero is partly consumed there. Delay the dose until the dormant period, as Hsu's team found, and more of it stays where it keeps grains apart. This is the same logic as casein mopping up tea catechins: whatever binds first wins.
Use this to check what you just read: plug in your own concrete mix and your own chai pan.
Order matters: the bucket of water and the ginger window
How far does the chai and concrete link really hold?
The link holds at the level of mechanism and breaks at the level of chemistry. Both systems contain a short, fast, irreversible window right after mixing, and in both the result depends on what is present during that window. The molecules, temperatures and time scales are completely different.
- Where it holds. Enzyme denaturation in chai and the first minutes of cement hydration are both fast first steps that consume or destroy something. Casein binding catechins and cement surfaces binding plasticiser are both binding problems of the Langmuir kind. Extra water dilutes the outcome in both, though only concrete punishes it with a law as sharp as Abrams'.
- Where it breaks. Chai is finished in five minutes and eaten; concrete gains strength for 28 days and beyond. Milk's split is a protein gel; concrete's strength is a mineral network of calcium silicate hydrate. Chai has no equivalent of curing, and no amount of order fixes a chai made with too little leaf.
- The temperature twist. The RSC's milk-first rule and Orwell's tea-first rule end at the same temperature. Pour 150 ml of tea at 95°C onto 30 ml of milk at 5°C and the mix settles near 80°C either way. Only the path differs, which is exactly what path dependence means.
The final recipe is the same; the history is not. In a pan or a mixer drum, the history wins.From the working above
Chai and concrete genuinely share a rule: a fast, irreversible step right after mixing acts on whatever is present first. Ginger boiled before milk loses the protease that otherwise has about 90 seconds in its 40 to 70°C window to start a curd; a plasticiser added 10 to 15 minutes late is not swallowed by cement's earliest reactions. Late water is punished in both, and in concrete by a precise law: about 27% of strength lost for each 0.1 of water-cement ratio. The analogy stops at the chemistry, which is protein in one and mineral in the other.

What is the best order for masala chai?
For a strong, smooth masala chai the sequence that the numbers support is water and spices, then leaf, then milk, then a final boil.
- Water, crushed ginger and whole spices. Boil 2 to 3 minutes. The ginger protease is destroyed and spice aromas extract into plain water.
- Tea leaf. Boil 1 to 2 minutes. Caffeine and catechins come out fast before casein is present.
- Milk. Add it and bring the whole pan back to a rolling boil. The casein now softens the astringency rather than competing for the extraction.
- Sugar. Any time; it dissolves and does not react on this time scale.
The full step-by-step recipe is in how to brew masala chai, and the curdling problem is answered in depth in why milk curdles in chai with ginger. How CTC leaf is made, which explains why chai leaf extracts so fast, is in CTC versus orthodox tea, and what ends up in your cup is ranked in caffeine in Indian drinks. For the story of how chai became India's drink, the Whyblog has the Tea Board's chai campaign.
What's new
- A minute-by-minute study of how ginger juice curdles hot milk Journal of Zhejiang University Science B, Apr 2025
- Lessons from deploying low-carbon LC3 cement in India RMI, Apr 2026
- From a Swiss lab to the Jewar airport runway: the cement that cuts emissions by 40% Business Today, Jul 2026
- JK Lakshmi Cement launches an LC3 product for low-carbon projects S&P Global, Feb 2026
What if next?
- A mango and a motorcycle both get better with heat, to a point. Enzymes and engine oil each have a temperature window, and both punish you outside it.
- The dabba in your bag is a thermos is a spacesuit. Layers, trapped air and shiny surfaces, from lunch to low Earth orbit.
- Tadka is a chemistry extraction, and so is perfume. Hot oil and a perfumer's lard obey the same distribution law.
Questions people ask
Should you add milk or water first when making chai?
Water first gives a stronger, smoother cup. Boiling ginger and spices in water destroys ginger's milk-clotting enzyme, and tea leaf releases most of its caffeine and catechins within about 2 minutes before casein can bind them. Add milk after and boil again.
Why does my chai curdle when I add ginger?
Ginger contains proteases that cut κ-casein like rennet. They are most active at 40 to 70°C, and in a 2025 study curd started forming about 90 seconds after ginger juice met hot milk. Boil the ginger in water before adding milk, or add it only to boiling milk.
Does adding water to concrete make it weaker?
Yes. By Abrams' law, strength falls exponentially with the water-cement ratio: each extra 0.1 costs about 27%. Ten litres added to a cubic metre with 350 kg of cement cuts strength by about 9%.
Is milk first or tea first better for British-style tea?
The Royal Society of Chemistry said milk first in 2003, because milk proteins degrade above about 75°C and drops of milk poured into hot tea briefly exceed that. Orwell said tea first in 1946, to control the amount of milk. Both end at about the same final temperature.
Sources
- Abrams DA: Design of Concrete Mixtures, Bulletin 1, Lewis Institute (1918, reprinted 1919) (Abrams' law, A = 14,000 psi, B = 8.2, about 50,000 tests)
- Hsu KC et al.: Effect of addition time of a superplasticizer on cement adsorption and on concrete workability, Cement and Concrete Composites (1999) (10–15 min optimum, Langmuir adsorption)
- Huang XW et al.: Purification, characterization, and milk coagulating properties of ginger proteases, Journal of Dairy Science (2011) (40–60°C proteolysis optimum, 70°C clotting maximum)
- Pan H et al.: Ginger-juice milk curd dynamics, Journal of Zhejiang University Science B (2025) (curd onset at 90 s, activity peak in first minute)
- Fernando CD, Soysa P: Extraction kinetics of phytochemicals during black tea brewing, Nutrition Journal (2015) (caffeine and catechins within 2 min, polyphenols at 6–8 min)
- Lorenz M et al.: Addition of milk prevents vascular protective effects of tea, European Heart Journal (2007) (16 volunteers, caseins complex with catechins)
- Royal Society of Chemistry: How to make a perfect cup of tea, press release text (2003) (milk first, denaturation above 75°C)
- NZ Herald: Science solves issue of milk or tea first (2003) (Andrew Stapley, Loughborough University, Orwell centenary)
- The Orwell Foundation: George Orwell, A Nice Cup of Tea (1946) (eleven rules, tea first, Indian tea)
- Wikipedia: Water–cement ratio (0.30–0.35 for full hydration, typical 0.40–0.60)
- Wikipedia: Duff Abrams (Lewis Institute, slump cone, 1918 publication)
- Wikipedia: Langmuir adsorption model (adsorption isotherm equation)

