Wine pairing: the six rules that actually work

Six things decide whether a wine pairing works, and every one of them has been measured in a laboratory. Colour is not among them.

Wine pairing is the management of six variables. Fat and protein blunt tannin, so a marbled cut or a hard cheese carries a wine that would strip a lean plate. Sugar in the dish and sugar in the glass suppress each other, so a dry wine served with dessert gives up its fruit and keeps its acid. Salt suppresses bitterness, which makes seasoning the fastest correction available at the table. Umami works by synergy rather than by weight, which is why champagne and oysters became a cliché. Iron, not colour, is what turns some reds fishy next to seafood. And alcohol amplifies chilli heat instead of quenching it, because ethanol acts on the receptor capsaicin acts on. Get those six right and the rest of what you have been taught is regional habit: worth keeping, and not worth obeying when it collides with one of the six.

Why do most wine pairing rules fail?

Because a pairing is not the sum of its parts, and the people who measure pairings say so in plain language.

Galmarini and colleagues at the Centre des Sciences du Goût et de l'Alimentation in Dijon gave 60 consumers three wines and three cheeses, then all nine combinations of them. The wines were a Pouilly-Loché Chardonnay, a Maranges Pinot Noir carrying 2,046 mg/L of tannin and a Beaujolais Gamay carrying 1,420 mg/L. The cheeses were Comté aged 14 months, Époisses and Chaource. Their conclusion on rules of thumb is blunt: the perception of a combination of products is not the result of an additive or subtracting effect which can be predicted based on their individual perception.

Then the result that should settle an argument you have had at a dinner table. The wine effect on liking was significant, and the white wine's combinations scored higher than either red's whatever cheese was on the plate, according to Galmarini and colleagues. In the authors' words, "white wine was a better companion for the evaluated cheeses than the red wines", and the reason was mechanical: "a reduced duration of astringency or bitterness as dominant".

Three things and only three drove liking down across those nine combinations, according to Galmarini and colleagues: sourness in seven of them, bitterness in six, astringency in five. The positive drivers were scattered across ten different descriptors. So a good pairing is mostly one that takes something away, and the six rules below are six ways of taking it away.

Rule	What is actually being measured	The move
Fat and protein blunt tannin	Perceived astringency of grape tannin against dietary lipids	Save the firmest wine for the richest plate
The glass must be sweeter than the plate	Mutual suppression of sucrose sweetness and citric acid sourness	Match or exceed the dish's sugar
Salt suppresses bitterness	Sodium salts against quinine, caffeine, magnesium sulfate	Season the food before you blame the wine
Umami needs synergy, not weight	Free glutamate against 5'-nucleotides	Put a glutamate wine with a nucleotide food
Iron drives fishy aftertaste	Total iron and ferrous ion against fishy aftertaste intensity	Ask about iron, not colour
Alcohol amplifies chilli heat	Ethanol lowering the heat-activation threshold of the capsaicin receptor	Drop the alcohol, not the flavour

Rule one: how much tannin can the plate carry?

As much fat and protein as the plate is carrying, and the gap between two bottles of one variety is wider than the gap between two varieties.

Saad and colleagues published the mechanism in the Journal of Agricultural and Food Chemistry in 2021. Catechin, "a majority component of grape tannins", interacts with the lipid droplets in an oil-in-water emulsion: the droplets grow, their growth then slows, and lipid movement rises in the layer at the droplet surface. Sensory analysis confirmed the consequence, according to Saad and colleagues: "dietary oils decrease the perception of astringency of grape tannin solutions". Their summary is worth keeping: "dietary lipids are crucial molecular agents impacting our sensory perception during wine consumption".

Cheese does the same job, and the size of the effect is smaller than most sommeliers imply. Madrigal-Galán and Heymann ran 11 trained judges over eight red wines and eight cheeses for the American Journal of Enology and Viticulture in 2006. Astringency fell. So did bell pepper and oak flavour. Butter aroma was the only attribute cheese enhanced, according to Madrigal-Galán and Heymann. But the same authors record that the sensory profiles with and without cheese were "very similar". Galmarini's 2016 study of 31 French consumers points the same way: in the Madiran, all four cheeses cut how long astringency and sourness stayed dominant, and red fruit aroma lasted longer.

Now the number that decides how much tannin you have to place. Harbertson and colleagues put 1,325 commercial red wines through a protein precipitation assay in 2008 and reported means of 672 mg/L catechin equivalents for Cabernet Sauvignon, 652 for Zinfandel, 559 for Merlot, 455 for Syrah and 348 for Pinot Noir, inside an overall range of 30 to 1,895 mg/L. Read the means and stop reading, and you will get this wrong. Harbertson reports a 32-fold spread inside Cabernet Sauvignon, and another inside Pinot Noir. Variety tells you the neighbourhood. It does not tell you the house. Two Cabernets can sit further apart than the 324 mg/L between the Cabernet and Pinot Noir averages, which is why a ripe Napa bottling like Opus One and a lean claret from an estate on the Bordeaux producer atlas belong on different plates.

The long version by protein is in best wine with steak, cut by cut and pairing wine with cheese.

Rule two: is the glass sweeter than the plate?

It has to be, or the wine hands over its fruit and keeps its acid.

Mao and colleagues measured the exchange in npj Science of Food in 2022 using sucrose and citric acid. Citric acid raised the detection threshold for sucrose from 0.424% to 0.624% and pushed the Weber fraction for sweetness from 20.5% to 33.0%, so sweetness became both harder to detect and harder to tell apart. Sucrose raised the detection threshold for citric acid from 0.0057% to 0.0082%, according to Mao and colleagues. That was a model solution rather than a wine, so treat it as the mechanism rather than the dose. The direction is the point: sugar on the plate spends your wine's sweetness, and what remains is acid and tannin standing on their own.

The labelling law shows how easily this gets misread. Under Annex III of Commission Delegated Regulation (EU) 2019/33, a still wine is "dry" up to 4 grams of sugar per litre, "medium dry" up to 12, and "sweet" only at 45 and above. Sparkling wine runs on its own scale, according to Annex III: "dry" means 17 to 32 grams per litre, and "medium dry" means 32 to 50. So the word dry on a bottle of fizz permits more sugar than the words medium dry on a bottle of still wine, and neither is anywhere near a dessert.

Which is why the classic dessert answers are not simply sweet wines but wines whose sugar is measured in tens of grams and carried on high acid. Château d'Yquem and its neighbours on the Sauternes producer atlas sit at that end of the scale. Put a dry white against a fruit tart and you have not made a pairing, you have made the wine taste sour.

Rule three: what does salt do to a tannic red?

It takes bitterness out of the equation, and it costs nothing.

The Institute of Medicine's 2010 review of sodium sets out the evidence, drawing on Breslin and Beauchamp: "various sodium-containing ingredients have been known to reduce the bitterness of certain compounds found in foods, including quinine hydrochloride, caffeine, magnesium sulfate, and potassium chloride". The same review records the second-order effect. Sodium acetate raised perceived sweetness in a mixture of sugar and the bitter compound urea by suppressing the urea and releasing the sweetness underneath, and it produced no such lift in a sugar solution that had no bitterness in it to remove.

Note what was tested. Those are pharmaceutical and mineral bitterants, not wine phenolics, so the step across to a young Left Bank claret is reasoning rather than measurement. The pairing evidence points the same way. In the Dijon study, according to Galmarini and colleagues, saltiness was a negative driver of liking for 78% of the panel when the Époisses was eaten on its own. It then turned into a positive driver in two of the combinations with the Maranges, cited by 65% and 83%. Salt that was too much on its own became the thing that made the pair work.

Practical version: before you decide a wine is too bitter for the dish, salt the dish properly and taste again. Ten seconds, no cost.

Rule four: when does umami actually help?

When the two sides bring different molecules. Umami is the one taste that multiplies rather than adds.

Schmidt, Olsen and Mouritsen took the most quoted pairing in the world apart in Scientific Reports in 2020. Free glutamate across the champagnes ran 1.4 to 7.5 mg per 100 mL, according to Schmidt, Olsen and Mouritsen, with the highest readings in older vintages that had spent 11 to 19 years on lees. That sits well under the 29 to 30 mg per 100 mL at which glutamate reads as umami on its own. The champagne cannot taste of umami by itself.

The oyster supplies the other half. The European oyster Ostrea edulis carried 30.4 mg of inosinate, 17.6 mg of guanylate and 94.0 mg of adenylate per 100 g of solid tissue, against 15.1, 9.2 and 42.4 for the Pacific oyster Crassostrea gigas, plus about 60% more free glutamate at 256.6 mg per 100 g against 160.1, as reported by Schmidt, Olsen and Mouritsen. Glutamate on one side, 5'-nucleotides on the other, and the umami appears in the combination rather than in either half.

That is a rule you can use. It says the pairing improves with lees-aged champagne rather than with a bigger wine, so a long-aged bottling such as Dom Pérignon or a late-disgorged cuvée from a house on the Champagne producer atlas brings more to a plate of shellfish than a young, fruit-forward fizz. It also says the species on the plate is a variable, and the flat native oyster is the stronger partner of the two they measured.

Rule five: why does red wine taste fishy with seafood?

Iron. Not tannin, not colour, and not the fish.

Tamura and colleagues, publishing in the Journal of Agricultural and Food Chemistry in 2009, had tasters work through 38 red wines and 26 white wines alongside dried scallop. They found strong positive correlations between the intensity of fishy aftertaste and the concentration of both total iron and ferrous ion. The causal test came next: adding ferrous ion to a model wine raised the aftertaste, and chelating the ferrous ion in a red wine suppressed it.

The consequence is narrower than the rule you were taught, and more useful. Red wine is not disqualified from a fish course, and the researchers themselves suggested low-iron reds might match seafood, as reported by MercoPress. What disqualifies a bottle is its iron, which no label states and which you cannot read off the colour. So treat "white with fish" as a working proxy that is wrong at the edges. The usual candidate for the exception is a light Pinot Noir, a village bottling from Vosne-Romanée, the commune behind Romanée-Conti, rather than a young Cabernet. Lightness is not a measurement of iron, though, and that is the honest limit of this rule: it stays a bottle-by-bottle question.

Rule six: what should you drink with chilli heat?

Something with less alcohol than you think, because alcohol works on the receptor the chilli works on.

Trevisani and colleagues reported the mechanism in Nature Neuroscience in 2002: "Ethanol potentiated the response of VR1 to capsaicin, protons and heat and lowered the threshold for heat activation of VR1 from approximately 42 degrees C to". VR1, now called TRPV1, is the receptor that produces the burn of chilli. Ethanol does not sit alongside capsaicin. It sensitises the channel capsaicin opens, and 34 °C is below the temperature of your own mouth.

The relief side has been measured too. Nolden, Lenart and Hayes gave 72 people spicy tomato juice and seven chilled drinks for Physiology & Behavior in 2019, rating burn every ten seconds for two minutes. Whole milk, skim milk and cherry Kool-Aid produced the largest reductions against water, according to Nolden, Lenart and Hayes. Whole and skim milk never differed at any timepoint, which led the authors to conclude that "milk is the best choice to mitigate burn, regardless of fat context, suggesting the presence of protein may be more relevant than lipid content". Cola and seltzer did less. They did not test wine.

Put the two findings together and the traditional answer to a hot dish holds up: lower alcohol, and enough residual sugar to cover the sugar in the sauce, which rule two already told you to check. The mistake is reaching for a big red because the food is loud. It is the alcohol in that bottle, not its flavour, that turns the burn up.

Does the bottle matter more than the grape?

Yes, and this is where a rule stops being enough.

Harbertson's 32-fold spread inside a single variety is the problem in one number. "Cabernet Sauvignon" tells you nothing about where a given bottle sits on the tannin ladder, and the label will not tell you either. Which band a wine is in, and whether its tannin still reads as primary, is a per-bottle question. It is also the part of pairing that a printed chart has never been able to answer, because a chart has one row per grape and the variation lives underneath that row.

If the bottle is already in your cellar, the drink-now shortlist answers the timing half of the question before the food question arises.

Get the matrix instead of memorising the rules

The six rules are the reasoning. What you want in a shop or at a table is the answer.

We keep it as a style pairing matrix: every style we score, its tannin, sugar and alcohol bands, and the plates each one holds up to, with the wine and producer pages sitting behind every cell. Subscribers get it as a one-page reference, so the decision takes ten seconds rather than a paragraph of theory you half remember. It updates as the scores do.

Start from the wine you own with red wine food pairing, or from the plate in front of you with white wine food pairing.