Despite the apparent simplicity of its ingredient list — pasta, Pecorino Romano, Tellicherry pepper — the Roman dish cacio e pepe is the subject of a disproportionate number of culinary failures. We present a systematic study of the underlying Pecorino-Starch-Water (PSW) emulsion, identifying the thermodynamic and mechanical conditions necessary for its stable formation.
Using a combination of dynamic shear rheometry, thermogravimetric analysis, and household-grade observation, we establish that (i) the stable PSW phase exists inside a narrow window bounded by 56 °C ≤ T ≤ 68 °C and ρstarch ≥ 0.028 g·mL−1; (ii) outside this window the system collapses into one of four failure modes; and (iii) the use of pre-salted pasta water and off-heat incorporation reduces failure probability by a factor of 4.3 ± 0.7 at the 95 % confidence level.
We conclude by offering a reproducible home-kitchen protocol calibrated against 412 consecutive in-lab attempts.
Cacio e pepe is, in its written recipe, the shortest pasta preparation in the Italian canon1. Its three ingredients — dry pasta, a hard aged sheep's-milk cheese (Pecorino Romano DOP), and black pepper — are assembled without cream, butter, oil, or any intervening emulsifier2. The physical result, when successful, is a single-phase suspension in which gelatinised pasta starch binds the aqueous calcium caseinate colloid into a self-stable, low-shear, creamy layer that coats the pasta surface uniformly at macaroni-scale resolution.
And yet.
Reports of unsuccessful outcomes outnumber successful ones in the domestic literature3,4. In the casual mode, the operator reports three symptoms: (a) clumping, the aggregation of cheese into macroscopic rubber beads; (b) slickness, a watery oleaginous film without cohesion; and (c) drag, a pasty residue that clings to the pan rather than the pasta5. Reviewing 412 in-lab attempts†, we document a fourth mode we term curdled dry, in which the cheese forms a crisp wafer immediately upon contact with residual pan heat6. The four modes correspond to four distinct thermodynamic excursions from the stable PSW region (Section 3, Figure 1).
In this paper we (i) quantify the boundaries of that stable region; (ii) identify the two operational parameters most sensitive to error; and (iii) propose a protocol whose success rate in a replication cohort of 24 participants exceeds 89 %.
All trials used Pecorino Romano DOP (24-month aged, Fratelli Masini, lot PR-2025-7A) grated on a Microplane 46020 and rested 12 min at 18 °C prior to use. Tellicherry black pepper (Vengalam, lot TP-2025-K14) was toasted in a dry pan for 40 s at 140 °C and crushed in a Peugeot U'Select to setting 3 of 6. Pasta was De Cecco tonnarelli no. 2 (Ø 2.00 ± 0.05 mm).
Bulk water temperature was monitored with a Thermapen IR ONE (±0.4 °C). Shear viscosity of the PSW phase was characterised using an Anton Paar MCR 302 rheometer (parallel plate, ∅ 50 mm, gap 0.5 mm, 25 °C) and complementary qualitative shear was assessed by one of the authors (M. T.) stirring the pan.
Critically, the cooking water was under-salted relative to standard practice — 7 g·L−1 NaCl versus the Roman convention of 10. This reduces Pecorino's solubility at the emulsion stage and counteracts the cheese's native salinity (≈ 6 g·100 g−1). Pasta was cooked for tcook = 4:30, i.e. 90 s short of the box time, to preserve surface starch.
where kleach is the empirically determined starch-leaching rate (kleach ≈ 1.3 × 10−3 g·g−1·min−1 for durum tonnarelli). For the protocol below, ρstarch = 0.031 g·mL−1, comfortably above the stability threshold.
The operator proceeds as follows. Pasta is cooked as in §2.3. Meanwhile, the pan (cold) is assembled with 60 g grated Pecorino and 30 mL of hot pasta water taken from the surface (preferential for starch concentration). A paste of approximately peanut-butter viscosity is formed by stirring at 200 rpm with a wooden spoon.
The pan is not placed on the heat. The drained pasta — still dripping, temperature Tpasta ≈ 84 °C — is added directly to the cold paste and agitated vigorously. Subsequent additions of pasta water are made in 10 mL increments; each addition must be fully incorporated before the next.
The final bulk temperature settles at 63 ± 2 °C: inside the stability window of Fig. 1. Freshly crushed pepper is folded in at this stage, not earlier, to preserve volatile sesquiterpenes7.
Equation (2) neglects the latent contribution of pan and utensil; for standard 28 cm carbon-steel, the thermal mass error is < 2 °C, well within the stability window.
We catalogue the observed outcomes below. A single mode typically dominates each failed trial, though combinations are possible under extreme conditions (e.g. grating with a box grater plus on-heat incorporation yields simultaneous clumping and curdled dry).
| Mode | Freq. | Symptoms | Root cause |
|---|---|---|---|
| Clumping | 34 % | Gummy, macroscopic beads of cheese; pasta naked | Pan above 74 °C at moment of cheese addition |
| Slick | 28 % | Watery film, no cohesion, pooling in plate | Starch below 0.020 g·mL−1 |
| Drag | 19 % | Paste adheres to pan rather than pasta | Too little water; overgrated cheese |
| Curdled dry | 14 % | Crisp wafer on pan floor; unsalvageable | Residual heat ≥ 82 °C + low water |
| Successful | 52 % | Uniform coat, glossy, no pooling, audible "crack" from the pepper | Window respected; cf. protocol |
Using a Latin-hypercube sampling approach across five parameters (pan temp, water:cheese ratio, grater fineness, salt content, incorporation rate), we find the sensitivity hierarchy to be:
where S denotes the binary success outcome (0, 1). In practical terms, the temperature of the pan at the moment of cheese incorporation is the single highest-leverage parameter and the one operators are least likely to monitor8.
| Component | Mass / Vol. | Notes |
|---|---|---|
| Tonnarelli, dry | 180 g | De Cecco n. 2 · Ø 2.00 mm |
| Pecorino Romano | 90 g | 24-mo DOP · Microplane #46020 |
| Tellicherry pepper | 2.2 g | toasted 40 s · crushed #3 of 6 |
| Pasta water (cook) | 1.6 L | 7 g·L−1 NaCl |
| Pasta water (reserve) | ~ 180 mL | scooped from surface late |
| Pan | 28 cm | carbon-steel · mpan 1.3 kg |
Why should a dish of three ingredients be so fragile? Pecorino Romano is unusual among cheeses in its exceptionally low moisture content (≈ 32 %) and high salt concentration. In a hot aqueous environment the casein micelles denature rapidly; in the absence of a stabilising colloid, they flocculate9. The stabilising colloid, in our case, is gelatinised starch: the amylose-rich oligomers leached from the pasta during cooking form a viscoelastic network whose storage modulus G′ is of the correct order to suspend the aggregating casein10.
The narrow 56–68 °C window is thus not a culinary quirk but a thermodynamic corridor: below 56 °C, the starch fails to gelatinise sufficiently; above 68 °C, casein denaturation outpaces the starch's ability to suspend it, and macroscopic rubbery aggregates form. The corridor's width (~12 °C) is unusual — for comparison, a crème anglaise is reliable across roughly 8 °C, a Hollandaise across 6.
The operator's instinct, upon seeing a sluggish sauce, is to increase the heat — precisely the worst action in this context. The paper's central pedagogical claim is that this instinct be replaced with the opposite one: remove heat, add a splash of pasta water, agitate.
Cacio e pepe fails reliably when operated as an emulsion and succeeds reliably when operated as a thermal puzzle. The four failure modes we document — clumping, slick, drag, curdled dry — each correspond to a boundary crossing in a single phase diagram (Fig. 1). A simple two-line recipe conceals a 12 °C corridor which, once respected, reduces domestic failure rates from approximately 1 in 2 to less than 1 in 9.
Future work will address (i) the analogous phase diagram of cacio, pepe e ricotta di bufala; (ii) altitude correction for the corridor width at reduced boiling points; and (iii) extension of the protocol to gluten-free tonnarelli surrogates, where kleach is nearly three-fold lower.
The authors thank Signora E. Rinaldi (Via del Pigneto, Roma) for the loan of her nonna's carbon-steel pan, and P. D'Aquila (Trattoria Da Felice) for permitting observational fieldwork during service. Funding: Consorzio Pecorino Romano (grant CPR-2024-03). The funding body had no role in study design, data collection, or the decision to publish.
G. P. maintains a personal affiliation with a family-owned Roman trattoria. The remaining authors declare no competing interests.
Raw rheometer traces, trial coding, and the Python notebook used for Fig. 1 are available at github.com/ gastr-lab/ cacio-e-pepe-412. Physical leftovers were consumed at the end of each experimental session and cannot be re-analysed.
If the dish clumps, you are on heat. If it slicks, increase starch (drop in a teaspoon of pasta water). If it drags, add 15 mL more pasta water and keep agitating. If it dries to a wafer, retire to the bedroom and order pizza.
ERRATUM v2 — pagination corrected 16 APR 2026. Table 1 row 4 previously listed "Curdled wet" — no such mode exists.