Milk
You cannot understand cheese without understanding the thing it’s made of, and milk is stranger than it looks. It appears to be a white liquid. It is actually a suspension of protein particles and fat droplets in sugar water, held stable by a mechanism that cheesemaking exists to break.
Everything on the family tree follows from what’s in here.
What’s actually in it
Roughly, for cow’s milk:
| Component | ~% | What it does in cheese |
|---|---|---|
| Water | 87 | The thing you spend the whole process removing |
| Fat | 3.5–5 | Flavour, texture, mouthfeel. Mostly trapped, not bonded |
| Protein | 3.2–3.5 | ~80% casein, ~20% whey protein. The casein is the cheese |
| Lactose | 4.8 | Bacteria food. Becomes the acid that drives everything |
| Minerals | 0.7 | Especially calcium phosphate, which is structural, not nutritional trivia |
The two numbers that matter most for a cheesemaker are the fat and protein, because together they set your yield — which is why the same recipe gives a bigger wheel from Jersey milk than from Holstein, and a bigger one in some seasons than others.
Casein: the part that becomes cheese
About 80% of milk protein is casein, and it doesn’t float about as individual molecules. It’s assembled into particles called micelles — and how those are built is the whole story of why milk is a stable liquid at all.
🔴 The picture in older books is wrong
If you learned this from a book more than about fifteen years old — and several on my shelf qualify — you were probably shown the submicelle model: the micelle drawn as a raspberry, a cluster of smaller sub-units packed together.
That model has been superseded. Researchers went looking for that particulate internal structure and could not find it. The current preferred description is quite different:
- a homogeneous matrix of casein polymers, not a bag of subunits;
- caseins forming linear and branched chains, two to five proteins long;
- those chains interlocked by nanoclusters of colloidal calcium phosphate — tiny, about 2 nm across, sitting at an average spacing of roughly 18.6 nm, described memorably in the literature as “cherry stones”;
- the whole thing an open, sponge-like supramolecule that resists being squashed or pulled apart.
I like this correction because the new picture is more useful at the bench than the old one. A raspberry of subunits suggests something that comes apart into pieces. A sponge held together by mineral staples tells you immediately why calcium matters so much, and why removing calcium weakens your curd. Which is exactly what happens when you pasteurise.
Why milk doesn’t curdle on its own
The micelles are covered in κ-casein, one of the four caseins, and it’s the one that behaves differently. Part of it sticks out into the surrounding liquid — a hydrophilic, negatively charged tail that projects from the surface like a hair.
Every micelle is wearing this fuzzy coat, and it keeps them apart two ways at once: electrostatic repulsion, because the tails are charged alike, and steric repulsion — the physical bulk of the protruding hairs simply preventing two micelles from getting close enough to bond.
That’s the stability cheesemaking has to defeat. Coagulation is entirely about removing that coat, and there are two completely different ways to do it.
Whey protein: the 20% that mostly gets away
The other fifth of the protein is whey protein, and it behaves nothing like casein.
It is not in the micelles. It stays dissolved. And critically, neither acid nor rennet coagulates it under normal cheesemaking conditions — so in an ordinary cheddar or gouda, that protein drains away with the whey and is simply lost from the wheel.
Two numbers explain how you get it back:
- whey proteins have an isoelectric point of about pH 5.1–5.3 — different from casein’s;
- they have poor thermal stability and denature above about 65 °C.
Heat them past that and they unravel, exposing sticky regions that bond to each other and to casein. That is the entire basis of the heat-acid cheeses — and it’s why true ricotta can be made from whey that has already given up its casein.
Fat: trapped, not bound
Milk fat travels as globules, each wrapped in a membrane. In cheesemaking, fat is essentially a passenger: the casein network forms around it and physically traps it.
Two consequences worth knowing.
Fat loss is real and visible. Handle fresh curd roughly and you tear the network before it’s knit; fat escapes into the whey, and you can see it — the whey goes cloudy and white rather than yellowish-green. That is yield leaving the vat, and you will not get it back.
And fat is where most flavour lives. Ripening works on fats as well as proteins, and the compounds that make an aged cheese taste of anything much are largely fat-derived. It’s why skimmed-milk cheese is not merely leaner but duller. (The ripening page is being written now.)
The variables you can’t see
Two things about the milk itself change your results more than most recipes admit:
Breed and season. Fat and protein percentages differ by breed, by stage of lactation, and by what the animals are eating. Spring grass milk is a different raw material from late-lactation winter milk. This is not a rounding error — it’s why an alpine summer cheese is a distinct thing.
Pasteurisation moves the calcium. Heat shifts some soluble calcium into an insoluble form, which weakens the mineral staples holding those micelle chains together. The practical result is a softer, weaker rennet curd from shop milk than from raw. Adding calcium chloride puts some of it back, and that is a genuine correction rather than an optional extra — see how to read a cheese recipe.
Next: coagulation — the two entirely different ways of stripping that coat off, and why one gives you a precipitate and the other a network.
Sources
- The nanocluster model and the superseding of the submicelle model — a homogeneous casein matrix with ~2 nm calcium phosphate nanoclusters at ~18.6 nm spacing, caseins in interlocked chains of two to five: Supramolecular Structure of the Casein Micelle, Journal of Dairy Science and a quantitative calcium phosphate nanocluster model
- Casein isoelectric point pH 4.6; whey protein isoelectric point pH 5.1–5.3 and denaturation above 65 °C: University of Guelph, Dairy Science and Technology and Cheese Science Toolkit — coagulation
Comments (0)