Why does ouzo turn milky with water when whisky stays clear? The ouzo effect, the louche, and the water that controls it

Key takeaways

• The cloud comes out of the spirit, not out of the water. Anise spirits carry trans-anethole, an oil that dissolves freely in ethanol and barely at all in water. Drop the alcohol strength by adding water and the oil has nowhere to go, so it nucleates into droplets that scatter light. Physicists call it the ouzo effect, a term coined by Vitale and Katz in Langmuir, 2003.
• The numbers are stark. Water dissolves anethole at about 0.111 g/L at 25 °C (PubChem, Aquasol database), while EU law requires a pastis to contain 1.5–2.0 g/L of it (Regulation (EU) 2019/787, Annex I, category 26). More than an order of magnitude more oil than water can hold.
• No emulsifier is involved. The droplets are roughly one micron across and stabilise themselves, growing by Ostwald ripening rather than coalescing (Grillo, Colloids Surf. A, 2003; Sitnikova et al., Langmuir, 2005).
• Water temperature measurably changes the result. Across 15–30 °C, both the peak turbidity and the alcohol fraction at which it occurs shift with temperature (Bickel, Ellis and Resnick, ACS Omega, 2021). The ice-cold water of the absinthe ritual is not only theatre.
• Dissolved ions change it too. In a trans-anethole/ethanol/water system, droplet number density falls once NaCl passes 10⁻³ M — about 58 mg/L — and droplets are most numerous and most stable near pH 6–7 (Gao et al., arXiv preprint, 2025).
• Whisky does the same thing. Below roughly 46% ABV, long-chain esters in an unfiltered malt come out of solution the same way — which is why the industry spends money on chill filtration to stop it (Scotch Whisky).

Ouzo turns milky because it contains an oil — trans-anethole — that is soluble in alcohol and almost insoluble in water, and adding water strips away the alcohol that was holding it in solution. The oil separates instantly into micron-scale droplets that scatter light, an effect known as spontaneous emulsification or the ouzo effect. Whisky stays clear at normal drinking strengths because its hydrophobic congeners are present in far smaller quantities and come out of solution at a lower threshold — and because most whisky is filtered specifically to stop it happening.

What is actually happening when the glass goes white?

Nothing is dissolving. Something is coming undone.

An anise spirit is a three-component system: water, ethanol and a hydrophobic oil. Ethanol is the go-between — miscible with water in all proportions, and a solvent for anethole too, so at bottling strength the oil sits in a clear single phase. Add water and the ethanol fraction falls. Past a certain point the mixture crosses into a region of the phase diagram where one phase is no longer stable, and the oil comes out — not as a slick on the surface, but as a dense cloud of droplets throughout the liquid.

The 2024 study that mapped this system describes the sequence plainly: the oil “is insoluble in water but fully soluble in alcohol”, so when water is added the alcohol is depleted and the system undergoes “spontaneous emulsification” (Archer et al., Soft Matter, 2024). What makes it remarkable is that it happens with no surfactant, no agitation and no energy input. Vitale and Katz named the phenomenon after the drink and explained it as homogeneous liquid–liquid nucleation (Langmuir, 2003).

The droplets are small and stay that way for a surprisingly long time. Small-angle neutron scattering on pastis put them on the order of a micron, with size depending on the anethole-to-ethanol ratio and growing with time and temperature (Grillo, 2003). Dynamic light scattering showed growth is governed by Ostwald ripening — large droplets slowly eating small ones — with no coalescence observed (Sitnikova et al., 2005). That is why a louche holds for a whole drink instead of breaking in a minute.

How much oil is in there, and how little will water hold?

This is where the arithmetic becomes persuasive.

Anethole’s solubility in water is listed at 0.111 g/L at 25 °C (PubChem, citing the Aquasol database). Now look at what the law requires in the bottle. Under Regulation (EU) 2019/787, pastis must have “a minimum and maximum anethole level of 1,5 and 2 grams per litre”, Pastis de Marseille “between 1,9 and 2,1 grams per litre”, and sambuca “not less than 1 gram and not more than 2 grams per litre” (Annex I, categories 26, 27 and 36).

So a pastis holds thirteen to eighteen times more anethole than pure water can dissolve. Every bit of that margin is being carried by ethanol.

Work through a normal serve. Take a pastis at 40% ABV with 2 g/L of anethole and dilute it 1:5 with water. The finished drink is around 6.7% ABV and holds roughly 0.33 g/L of anethole — about three times what plain water can dissolve, and the residual ethanol is nowhere near enough to cover the gap. Most of that oil has to leave solution, and it does so as an emulsion rather than a slick because it nucleates everywhere at once.

For absinthe, “most absinthe drinkers use a water-to-absinthe ratio of between 3:1 and 5:1”, finishing the drink at roughly 11–15% ABV (Wormwood Society). The ritual exists because the louche is the signal that the dilution has gone far enough.

Why doesn’t whisky do this?

It does. It is simply held back, in two ways.

The hydrophobic congeners in malt whisky — long-chain fatty acid esters rather than a terpenoid oil — are present at milligrams per litre, not grams, and the threshold at which they precipitate sits far lower down the strength scale. The figure the trade quotes is around 46% ABV: above it those esters generally stay in solution at room temperature, below it they may not, and raising bottling strength to about 46% is the standard way to release a whisky non-chill-filtered and keep it bright (Scotch Whisky). The full mechanism is in why whisky goes cloudy when you add water.

So the difference between the two glasses is not physics. It is dose and intent. In an anise spirit the cloud is the product — what the customer is paying for, and what the producer formulates to guarantee. In whisky the same cloud is a cosmetic defect, and an entire unit operation exists to prevent it. One industry engineers the ouzo effect in; the other engineers it out. Nor are anise spirits alone: baijiu, shochu and others carry enough higher esters to haze on dilution, which is why water specification matters as much there as in Scotch (soju, shochu and baijiu).

Does the water you add actually change the cloud?

Yes — measurably, in at least two respects.

Temperature. The tradition of dripping iced water over absinthe has a physical basis. Bickel, Ellis and Resnick measured optical transmittance through absinthe as a function of alcohol fraction at five temperatures from 30 °C down to 15 °C, and found that both the maximum turbidity and the alcohol fraction at which that maximum occurs are temperature-dependent, with the fitted parameters varying linearly with temperature (ACS Omega, 2021). Grillo’s neutron scattering points the same way from the structural side: droplet size grows with temperature (2003). Colder water gives a different cloud from warmer water — not merely a colder drink. The same thermodynamics that makes proofing tanks warm up is at work here, as covered in heat of mixing and blending water temperature.

Dissolved ions. A 2025 preprint studying exactly the trans-anethole/ethanol/water system reports that as NaCl concentration “surpasses 10⁻³ M, the number density of nucleated mesoscopic droplets decreases significantly”, with nearly 90% of those droplets gone at 1 M, while mean droplet diameter climbs from around 150 nm to 600 nm (Gao et al., arXiv:2505.13835). The same work finds droplets most numerous and most stable near pH 6–7, where they carry their maximum negative surface charge, and reports rapid coarsening below pH 3 and near-total suppression at pH 12.

Keep the scale honest. That 10⁻³ M threshold is about 58 mg/L of sodium chloride, comparable to the dissolved solids of a soft tap water — the point at which the effect merely becomes detectable. The 90% loss occurs at 1 M, roughly 58 g/L, which is seawater territory. The reading is directional rather than alarming: ionic strength and pH are variables in this system, fewer ions and a near-neutral pH favour a finer and more populous emulsion, and a water of known composition gives a repeatable result while an unknown one does not. For how those numbers are expressed, see pH and TDS explained and water hardness units.

A third factor has nothing to do with the louche. Water added at the table is undiluted by anything else, so whatever it tastes of goes straight into the glass — chlorinated tap water most of all, since chlorine can generate chlorophenols detectable at parts per trillion (the chlorophenol problem). An anise spirit is aromatically assertive, but it is not immune, as whisky with water or ice argues at length.

What does this mean for the water that proofs an anise spirit in the first place?

Here the louche stops being a party trick and becomes a production constraint.

An anise spirit is usually built from high-strength neutral or grape spirit and reduced with water, exactly as a gin or vodka is. Reduce 96.4% ABV spirit to a 40% pastis and water is roughly 58% of the finished bottle — the water-heavy arithmetic set out in what is blending water and proofing water for gin, vodka, rum and tequila.

But there is an extra constraint no whisky bottler faces. The product must arrive clear in the bottle and louche only in the customer’s glass, so bottling strength has to sit far enough above the precipitation threshold to survive a cold warehouse, a winter container and a shop window. Since the alcohol fraction at which turbidity develops moves with temperature (Bickel et al., 2021), a product formulated close to its threshold at 20 °C is one that can go hazy at 5 °C. Three consequences follow:

• Proofing accuracy becomes a visual specification, not only a legal one. An overshoot on water addition that a whisky would shrug off can push an anise spirit towards permanent in-bottle haze. The arithmetic and its tolerances are in how much water do you actually add.
• Calcium and magnesium deserve tight limits, especially in sweetened styles. Sambuca carries a minimum of 350 g/L of sweetening products (Reg. (EU) 2019/787, Annex I, cat. 36), and sugar-rich, oil-rich, fatty-acid-bearing systems are exactly where hardness ions form insoluble soaps and throw a haze that never clears. That failure mode is set out in liqueur and cream liqueur water specification.
• Consistency of the water matters more than its character. Because ionic strength is one of the variables that shapes the emulsion, a water whose composition drifts seasonally gives a louche that drifts with it. A demineralised water is not “flavourless water” here; it is the one input whose contribution is reliably close to zero. The specification is in demineralised water specifications for spirit dilution, and the case against adding minerals back is in can water be too pure.

What should a producer or a buyer actually check?

Four things, in order of how often they go wrong.

1. Bottling strength margin. Establish the strength at which your own formulation louches, at the coldest temperature the product will realistically see, and set the bottling target with headroom above it. This is a formulation decision that the water then has to deliver reliably — and it interacts with the legal floor discussed in minimum bottling strength.
2. Conductivity and hardness on the blending water. Low, stable, and evidenced on every delivery rather than asserted once. See how blending water is tested.
3. pH. Near-neutral is where the emulsion literature puts maximum droplet stability (Gao et al., 2025). A demineralised water reading slightly acidic because it has absorbed atmospheric carbon dioxide is normal and unbuffered, not contaminated.
4. Cold stability testing. Chill the finished product, hold it, and look at it. It is the cheapest test in the building and the one that catches the failure the physics predicts.

And for anyone simply pouring a glass: use cold water, add it slowly, and stop when the cloud tells you to. The spirit was designed to signal its own dilution — which is more than whisky does, where you have to judge it by nose, as described in why water opens whisky.

Glossary

• Louche — the milky opalescence that appears when water is added to an anise spirit, and by extension the act of producing it. From the French for “cloudy” or “shady”.
• Ouzo effect — spontaneous emulsification of a hydrophobic oil when a water-miscible solvent holding it is diluted, named after the drink by Vitale and Katz in 2003.
• trans-Anethole — the principal aromatic compound of anise, star anise and fennel; a hydrophobic oil with a water solubility of about 0.111 g/L at 25 °C.
• Spontaneous emulsification — formation of an emulsion without mechanical energy or an added surfactant, driven purely by a change in solvent composition.
• Binodal and spinodal curves — the boundaries on a phase diagram between a stable single phase, a metastable region where droplets must nucleate, and an unstable region where the mixture separates outright. The ouzo region sits between them.
• Ostwald ripening — growth of larger droplets at the expense of smaller ones by diffusion of dissolved material, as opposed to droplets merging on contact.
• Chill filtration — chilling a spirit to around 0 °C and filtering it to remove compounds that would otherwise precipitate at low temperature or on dilution.
• Anise spirits — the family including pastis, ouzo, raki, arak, absinthe, sambuca, anisette and mastika, all flavoured with anethole-bearing botanicals.

Frequently asked questions

Does the louche mean the spirit has gone off? No. It is the expected and intended behaviour of the product, and it happens within seconds of adding water to a perfectly sound bottle. A haze that appears in the sealed bottle before any water is added is a different matter and usually points to a formulation or temperature problem.

Will demineralised water still produce a louche? Yes, and it is the cleanest way to produce one. The cloud is caused by the oil leaving solution as the alcohol fraction falls, which depends on how much water you add rather than on what is dissolved in it. Purified water simply removes the variables — ionic strength, hardness, chlorine taste — that would otherwise vary between one glass and the next.

Does hard water ruin a louche? Not at drinking-water concentrations, on the current evidence. Ionic strength does affect the system — droplet number density falls and droplet size rises as NaCl increases past 10⁻³ M, roughly 58 mg/L (Gao et al., 2025) — but the striking losses in that work occur at concentrations thousands of times higher than any potable water. Treat mineral content as a consistency variable, not a catastrophe.

Why is ice-cold water specified for absinthe? Partly sensory and partly physical. Turbidity and the alcohol fraction at which turbidity peaks both vary with temperature across the 15–30 °C range (Bickel et al., 2021), and droplet size grows with temperature (Grillo, 2003). Cold water gives a finer, denser cloud as well as a colder drink.

How much water should I add to ouzo or pastis? Less than absinthe needs. Absinthe is customarily taken at 3:1 to 5:1 water to spirit (Wormwood Society), finishing around 11–15% ABV; ouzo and raki are more often taken at roughly 1:1 or 2:1 with ice. Add slowly and let the louche tell you when you have arrived.

Can the same thing happen to my whisky? Yes, if it is non-chill-filtered and you take it below about 46% ABV, or add ice. The mechanism is the same class of phenomenon with different molecules at much lower concentrations, and it is cosmetic rather than a fault — see why whisky goes cloudy when you add water.

Is the ouzo effect useful for anything other than drinks? Very much so. Surfactant-free nanoemulsions made by exactly this route — dissolve a hydrophobic compound in a water-miscible solvent, then dilute — are a standard laboratory method for producing drug and polymer nanoparticles. The phase behaviour of the ouzo ternary system has been mapped precisely for that reason (Archer et al., Soft Matter, 2024).

Blenders International · Water Science series