Cacio e Pepe Recipe (the Roman Way, Without the Clumps)

What cacio e pepe is, and why three ingredients are harder than ten

Cacio e pepe is, literally, ‘cheese and pepper’: pasta tossed with finely grated Pecorino Romano, coarsely cracked black pepper and some of its own starchy cooking water until a creamy sauce forms and clings to the strands. That is the entire dish. There is no butter, no oil, no cream and no garlic in the Roman version — the creaminess is an emulsion the cook builds in the pan, and it exists for about two minutes before it begins to set.

It is also, by common Roman consent, the easiest of the city’s pastas to ruin. Carbonara gives you egg yolk — lecithin, a born emulsifier — to hold the sauce together. Gricia and amatriciana give you rendered guanciale fat. Cacio e pepe gives you nothing: cheese proteins, water and heat, with only the starch from the pasta water standing between you and a clumped, greasy failure. That is why a search for an easy version so often leads to a different dish — butter and cream make the emulsion nearly unbreakable, and they are not what the name describes.

This page follows the same doctrine as our carbonara: understand the emulsion and the dish becomes reliable. The levers here are the same three — temperature, starch, agitation — turned further, because there is no safety net under them.

Three ingredients, and what each one is really doing

Pecorino Romano is a hard, salty, sharp cheese made from whole sheep’s milk in a protected zone covering the whole of Lazio, the whole of Sardinia — where most of it is now made — and the Grosseto province of Tuscany, aged a minimum of five months (eight for grating wedges), and salted heavily along the way — it was provisioning food for shepherds and travelers long before it was a pasta cheese, and salt is how it survived. In this dish it is the sauce, the seasoning and most of the calories at once. Its proteins are also the failure mode: heat them too fast and they cross-link into elastic strings, squeezing out fat and water — which is exactly what a clump is. ‘Cacio’ is the old central-Italian word for cheese itself, from the same Latin caseus that gives us ‘casein’ — regional, not textbook, Italian — and in Rome the default cheese it meant was, always, pecorino.

The pepper is the second named ingredient, and it arrives in three registers. First, toasting: a minute or two in a dry pan wakes the volatile oils, and you will smell the difference immediately — resinous, citrusy, almost floral over the sharp bite. Second, the infusion: half the cracked pepper simmered in pasta water in the skillet, so pepper flavor is inside the sauce, not sitting on it. Third, the finish: coarse cracks over the plated nest, where they hit the nose before the fork moves. Buy whole peppercorns — pre-ground pepper lost its aromatics months ago and contributes only heat and grey dust — and crack them jagged, not fine: powder makes the sauce muddy, coarse cracks make it ivory with black constellations. Our 1 tbsp for four portions is a floor, not a ceiling; Roman versions often go further, and if you have a choice of peppercorns, a bright Tellicherry suits the cheese. The dish is meant to bite.

The pasta and its water are one ingredient in two states. The wheat starch that the pasta sheds as it cooks is the emulsion’s stabilizer: dissolved starch granules swell, burst and coat the cheese proteins, physically getting in the way of the cross-linking that makes clumps. This is why every serious recipe cooks the pasta in less water than usual — the same starch in half the water is twice the protection — and why the sauce is built from cloudy cooking water rather than tap water. Understand that and you understand the dish.

Why it clumps: the physics, including the 2025 paper

Pecorino clumps because its casein proteins aggregate. Somewhere past 140-150°F (60-65°C) they denature, knit to their neighbors and squeeze out fat and water — elastic wads in a thin, greasy puddle. Boiling pasta water is 212°F (100°C). A heap of grated cheese meeting a screaming-hot pan of it is a reliable way to make squeaky curds, and no amount of stirring will argue the proteins back apart. Starch is the bodyguard: dissolved starch raises viscosity and physically blocks protein-to-protein bonding, which is why cloudy, concentrated pasta water works and clear water does not. Agitation — the unphotogenic tossing — shears the cheese into the starchy liquid so the fat stays in small droplets and the proteins stay separate. Heat, starch, agitation: skip any one and the sauce breaks.

In 2025 a group of Italian physicists living abroad — who cooked together in their expat kitchens and, as one of them put it, had ‘the practical aim to avoid wasting good pecorino’ — published a peer-reviewed study of exactly this in the journal Physics of Fluids, mapping when and why the sauce collapses. Their phase diagram gives the failure a name: the ‘Mozzarella Phase’, the regime where the proteins fuse into stretchy clumps. Two variables drive the transition — temperature and starch concentration — and the paper’s central finding is that starch moves the cliff edge: with almost no starch in the water, the cheese collapses into clumps at around 150°F (65°C), while at the right starch level the same sauce shrugs off far higher heat. Their number for the right level: starch at about 2-3% of the cheese’s weight. Real pasta water is the gamble — its starch content is wildly variable, and a big pot of barely cloudy water may carry a fraction of what the sauce needs — which is why the same technique works one night and fails the next, and the cook gets the blame when the water was the problem.

Their recommendation for certainty: stop gambling and measure. Dissolve a small, weighed amount of potato starch or cornstarch in water, heat it gently until it thickens and turns nearly clear, and build the cheese cream on that. For the 7 oz (200 g) of cheese in this recipe, that is the 1 1/2 tsp (4-5 g) ‘insurance’ in the ingredient list. (The paper also names a second stabilizer — trisodium citrate, the melting salt behind processed cheese — for a perfectly smooth, slightly clinical sauce; and the whole enterprise won its authors the 2025 Ig Nobel Prize in Physics, the award for research that makes you laugh and then think.) You do not need a physics degree to make cacio e pepe, and Romans made it for generations without one — but the paper explains the dish’s reputation honestly: the traditional method has an invisible variable in it. The crema method in our steps narrows the gamble; the measured starch removes it. Use the insurance the first few times, then decide whether you still need it.

Cacio e pepe sauce: there is no jar

Cacio e pepe sauce is not a condiment you hold in the fridge; it is an emulsion that comes into existence in the pan in the last two minutes and begins dying the moment you stop tossing. Searches for ‘cacio e pepe sauce’ often want something pourable, something to bottle. What you can bottle is closer to alfredo — the jarred products sold under this name are often cream-based, alfredo-style sauces flavored with pecorino, and even the best-known cream-free jar holds together with cornstarch and sodium citrate, which are, delightfully, the exact two stabilizers the physics paper reached for. What you can make is pecorino, cracked pepper and starchy water forced into a single glossy phase, and it does not exist until the pasta is in the pan.

In practice the sauce has two moments. First, the crema: finely grated cheese worked with a little cooled pasta water into a paste, then loosened to heavy-cream thickness — already an emulsion, built below the temperature where the proteins panic. Second, the marriage in the skillet: underdone pasta, the pepper infusion, the crema and more starchy water, tossed off the heat until the strands are lacquered. If you have watched a Roman cook finish the dish in the pan, that is all that is happening — no secret fourth ingredient, just water and temperature controlled with the wrists.

When the sauce is right it clings, stretches slightly between strands, and looks wet without puddling. When it is short of water it turns to salty paste as it sits; when it is broken you will see either sandy grains or elastic clumps. Both are recoverable if you catch them early — the rescue protocol is at the bottom of this page — and both are why you plate immediately.

The methods, honestly ranked

The trattoria toss — cheese grated fine, pasta tossed with pepper and starchy water, cheese rained in off the heat while tossing hard — is the fastest and the most fragile. In a Roman kitchen like Felice a Testaccio’s, where the dish is finished with two forks at the table and the cooks make forty a night, it produces the best plate of the four methods. At home it is the highest-variance route: the pan is often still too hot, the water ratio is a guess, and there is no paste stage to absorb mistakes. It is authentic, and it is how most first attempts fail.

The crema method, which this page teaches, moves the risk off the stove: the cheese is worked into a smooth paste with warm-not-hot pasta water in a bowl, so hydration happens gently and completely before the cheese ever meets the pasta. The pan then only has to coat and loosen. It costs two minutes, forgives a lot, and it is still nothing but cheese, pepper, pasta and water. Most modern Roman restaurant kitchens work this way in some form; it is the method to learn first.

The measured-starch method is the crema method with the physics paper’s correction: a weighed teaspoon of potato starch, dissolved and heated, replaces the guess about what your pasta water contains. It is the most reliable route to the true no-fat Roman sauce, and for gluten-free pasta — whose water carries far less usable starch — it is effectively mandatory. Purists may bristle at starch from a box; the sauce cannot tell the difference, and neither, honestly, can the purists.

Butter — a knob mounted into the pan with the pasta water before the cheese goes in — is the American restaurant road, and it deserves a civil word rather than a sneer: butter is already an emulsion, its milk proteins and fat widen the temperature window almost unconditionally, and the result is glossy, rich and nearly foolproof. It is also a different dish — rounder, blunter, tilted toward cow rather than sheep — closer to a burro e formaggio wearing cacio e pepe’s name. If you are cooking for eight, or for the first time for guests, nobody at the table will file a complaint. Make it on purpose, and know what the Roman version tastes like so you know what you are trading. Cream goes further still and erases the point of the pecorino; save it for a sauce that wants it.

Tonnarelli, spaghetti, and what the sauce wants from a shape

In Rome the dish’s classic partner is tonnarelli — the Roman name for spaghetti alla chitarra, a fresh egg pasta cut square on the steel wires of a chitarra so each strand has four rough faces. Its porous, eggy surface drinks the sauce, and its extra chew stands up to the pepper. If you can buy or make it, it is the reference experience — with two caveats: fresh egg pasta sheds less starch into its water than dry semolina pasta (have the starch insurance ready), and it cooks in a few minutes, so the crema must be ready before the pasta hits the water.

Dry spaghetti is what most Romans cook at home and what this recipe assumes: choose a bronze-cut brand, whose matte, sandpapery surface holds the emulsion the way glossy Teflon-die pasta cannot, and whose semolina sheds generous starch. Rigatoni and mezze maniche work on the same logic as they do for carbonara. What the sauce cannot use is anything too thin — angel hair overcooks in the time the toss takes and cools so fast the cheese sets on contact — or fresh ribbons soft enough to tear under the vigorous tossing this dish demands.

A shepherd’s dish that was in the canon before carbonara existed

In the usual telling, cacio e pepe is the oldest of Rome’s four pastas: the shepherds of the Roman campagna and the Apennine transhumance routes walked for months with exactly this pantry — dried pasta, whole aged pecorino and black peppercorns, three foods that survive a saddlebag in summer, weigh little and carry enormous calories — and cheese and pepper over pasta is simply what they make when water is the only other thing available. Hold the story a little loosely: it is traditional attribution rather than documented history, gricia’s partisans tell an almost identical shepherd story and claim their dish as the elder, and pepper was a bought trade good, not something a flock provides. What a responsible page can date is the cookbook record.

And the record is clear on the order of things: cacio e pepe was already canon when the Roman repertoire was first written down — Ada Boni’s La cucina romana of 1930 includes it as a matter of course — while carbonara is absent from the record before 1939. The family tree we drew on the carbonara page starts here: cacio e pepe is the trunk. Add guanciale and you have gricia; add tomato to gricia and you have amatriciana; add egg to gricia and you have carbonara. Master this dish — the emulsion with no help — and the other three are additions, not new techniques.

Its recent history is a second act: in the 2010s cacio e pepe went from Roman comfort food to global restaurant obsession, turning up butter-mounted in New York, folded into cocktails and croissants, and — in the version that helped make Riccardo Camanini’s Lido 84 famous — cooked inside an inflated pig’s bladder presented at the table. The dish survived the fame intact at home: it still costs three ingredients and half an hour, and the trattoria plate in Testaccio has not changed.

Is cacio e pepe healthy? The honest ledger

A serving of this recipe — a quarter of 14 oz (400 g) of pasta and 1.75 oz (50 g) of Pecorino Romano — comes to roughly 570 calories: dried pasta at about 370 kcal per 100 g gives 370, the pecorino at about 390 kcal per 100 g gives just under 200, and the pepper is a rounding error. That is a genuine Roman primo portion, and it is lighter than our carbonara’s 770, for the simple reason that there is no pork fat, cream or butter in it — the entire fat load is the cheese’s own, roughly 15 g per serving, of which about 9 g is saturated.

Sodium is this dish’s honest problem — nearly as big as carbonara’s, despite containing no cured pork at all. Pecorino Romano is among the saltiest cheeses sold — about 1,400 mg of sodium per 100 g in USDA figures for domestic Romano, and up to nearly 2,000 for some imported DOP wedges — so the 50 g on each plate carries roughly 700-1,000 mg before the pasta water contributes its share. Call it 850-1,200 mg a serving with our deliberately under-salted water, and more if you salt the water normally: a third to half of the 2,300 mg daily ceiling in US dietary guidelines, in one primo. The recipe’s levers are already pulled — a third of the usual salt in the water, none anywhere else, and nothing extra at the table. The remaining lever is portion, which is why the Roman answer is a modest first course, not a main-course trough.

Do not try to lighten it by substitution. Low-fat hard cheese seizes faster and tastes of less; stretching the sauce with yogurt or cottage cheese makes a different, sadder dish; and cutting the cheese in half just makes weak, watery pasta. Eat the real thing, in the Roman portion, beside something green and bitter, and not every night.

Serving it: the two-minute window, the wine, the scaling trap — and the rescue

Everything about serving cacio e pepe is downstream of one fact: the sauce sets as it cools, and it is cooling from the moment the tossing stops. Bowls warmed with a splash of pasta water, the table already set, the diners already sitting — the Roman habits are not fussiness, they are engineering. Serve it as a primo in the Italian portion, with something sharp and bitter beside or after: puntarelle with anchovy dressing in winter if you can find it, otherwise chicory or arugula (rocket) with lemon. No bread; there is nothing to mop that the pasta has not already claimed. For wine, stay in Lazio and stay crisp: a Frascati or another dry Castelli Romani white against the salt, or — if you want red — a young, unoaked Cesanese served slightly cool. The salt is the loudest thing on the table, and tannin quarrels with it.

The scaling trap, stated plainly: this recipe does not double well. Eight portions means a pot of water so large the starch dilutes to nothing, a mass of pasta too heavy to toss, and a pan that cools unevenly — the pasta at the edges cold while the center scalds, so the crema meets two wrong temperatures at once. Four portions in one pan — this recipe — is the sweet spot; six is the ceiling, with the measured-starch method and a very wide pan. For eight, cook two batches back to back with the crema already mixed, or take the honest butter off-ramp, which scales without complaint.

And the rescue, for the night it breaks anyway. Sandy grains or the first elastic strings: pull the pan off the heat immediately, add a splash of cool or lukewarm pasta water — never boiling, which cooks the clumps into permanent rubber — and toss hard; you are dropping the temperature and putting shear and dilution back to work, and a borderline sauce will often come back. Escalation: work a fresh spoonful of powdered pecorino with cool water into a quick new crema and beat it in off the heat, or add a pinch of the starch slurry — the physics paper’s insight in miniature, more starch between the proteins. Fully seized, squeaky lumps you could pick out with a fork are past saving: fish them out, re-dress the strands with a fresh crema, and say nothing. Every Roman has been there.