What Blind Tasting Can Settle

The obvious way to settle this is to take the labels off.

Blind tasting looks like the referee both sides need, and it is a worse one than either wants, for reasons worth understanding rather than dismissing.

Here is what it reliably shows.

Experienced tasters place wines correctly above chance. Under controlled conditions, repeatedly, people who have spent years drinking Burgundy can put a glass on the right commune more often than random assignment would predict, and sometimes on the right vineyard.

That result matters enormously and the sceptical camp tends to skip past it. If somebody can identify a hillside from a glass of liquid, the hillside is physically present in the glass. There is nothing else in front of them. No amount of argument about suggestion and expectation gets around the fact that the information had to arrive somehow.

Here is what it also reliably shows.

Show the label and the score moves. Predictably, measurably, in the direction of expectation. The same wine scores higher from a famous bottle than from an anonymous one, and this holds for professionals, and it holds for professionals who know about the effect and are trying to compensate.

Tasters misidentify famous wines regularly. They disagree with each other on the same flight. They disagree with themselves on the same wine a week apart.

And underneath all of it sits the problem nobody can design around: bottle variation. Two bottles of the same wine, from the same case, stored together, can differ more than two adjacent vineyards do. Cork is a natural product and it is not consistent.

So what is the honest position?

Blind tasting is an instrument with a resolution, like any other, and the mistake both camps make is treating it as a verdict rather than as a measurement with a known error bar.

It reliably detects large differences. Gevrey against Chambolle, Burgundy against Oregon, a great vintage against a poor one: these are within its resolution and it gets them right far more often than chance.

It does not reliably detect small ones. Two premiers crus in the same commune from the same producer, two bottles of the same wine, the difference a slightly different picking date makes: these are below its resolution, and results at that scale are noise dressed as findings.

The trouble is that almost every argument people actually want to have is in the second category.

Nobody needs a referee to tell them that Chambolle and Gevrey differ. The disputes are about whether this premier cru deserves promotion, whether that grower's 2019 is better than their 2018, whether the extra money for the higher classification is justified in this specific case. Those are precisely the questions the instrument cannot answer.

So cite the successes and ignore the failures, or the reverse, and you can leave the room believing whatever you believed on the way in. Which is what both sides reliably do, and why the referee has never settled anything.

What it does establish, and this is not nothing, is that there is real signal in the glass and a real effect from the label, and that any honest account of wine has to contain both.