Republic of Mathematics blog

Tasting wine: a coin toss?

Posted by: Gary Ernest Davis on: April 14, 2011

A valuable feature of  quantitative reasoning – also known as numeracy, or quantitative literacy – is its ability to act as a BS detector.

John Allen Paulos (@JohnAllenPaulos) makes this clear in his book “A Mathematician Reads the Newspaper“, though expressed in a more gentlemanly way as saying mathematics can clarify many situations.

This morning I read an account in The Guardian of how people are no better at predicting whether a wine is cheap or expensive than they are at predicting a coin toss:

Expensive wine and cheap plonk taste the same to most people

Ian Sample, science correspondent

The Guardian, Thursday 14 April 2011

In a blind taste test, volunteers were unable to distinguish between expensive and cheap wine

People fool themselves into thinking expensive wines taste better than cheap ones, says psychologist Richard Wiseman. Photograph: Graham Turner/Guardian

An expensive wine may well have a full body, a delicate nose and good legs, but the odds are your brain will never know.

A survey of hundreds of drinkers found that on average people could tell good wine from plonk no more often than if they had simply guessed.

In the blind taste test, 578 people commented on a variety of red and white wines ranging from a £3.49 bottle of Claret to a £29.99 bottle of champagne. The researchers categorised inexpensive wines as costing £5 and less, while expensive bottles were £10 and more.

The study found that people correctly distinguished between cheap and expensive white wines only 53% of the time, and only 47% of the time for red wines. The overall result suggests a 50:50 chance of identifying a wine as expensive or cheap based on taste alone – the same odds as flipping a coin.

Richard Wiseman, a psychologist at Hertfordshire University, conducted the survey at the Edinburgh International Science Festival.

“People just could not tell the difference between cheap and expensive wine,” he said. “When you know the answer, you fool yourself into thinking you would be able to tell the difference, but most people simply can’t.”

All of the drinkers who took part in the survey were attending the science festival, but Wiseman claims the group was unlikely to be any worse at wine tasting than a cross-section of the general public.

“The real surprise is that the more expensive wines were double or three times the price of the cheaper ones. Normally when a product is that much more expensive, you would expect to be able to tell the difference,” Wiseman said.

People scored best when deciding between two bottles of Pinot Grigio, with 59% correctly deciding which was which. The Claret, which cost either £3.49 or £15.99, fooled most people with only 39% correctly identifying which they had tasted.

___________________________________________________________________________

So what does this article, and the study conducted by Professor Wiseman, say about people’s ability to distinguish wines?

The headline proclaims that “Expensive wine and cheap plonk taste the same to most people“, but is that what the study showed?

What we are able to gather is that about half the time people could not tell the difference between cheap and expensive wine.

Does this mean that each individual person might as well have tossed a coin to decide on wine quality?

Not at all.

Imagine that half the people tested were experts and got it right every time, while the other half were dunces who got it wrong every time.

On average, people would be right half the time.

Could this happen any other way?

Of course: half the people, the “experts”, might get it right 75% of the time, while the other half the people got it wrong 75% of the time.

Again, on average, people would be right half the time.

There are many ways that a heterogeneous group of people could come out getting the wine quality right half the time on average.

This says nothing about any individual person as the headline seems to suggest.

What the study does suggest is that if we randomly chose a person from the population tested and asked them to judge the quality of wine we would see a correct result about half the time.

That’s quite different to saying that if we choose an individual and test them repeatedly, they will be right about half the time.

 

 

 

 

3 Responses to "Tasting wine: a coin toss?"

So the study revealed that 50% of people could get the wine quality right. Not 50% of the time, people get the quality of the wine right.

You’re right – big difference.

But I’m not clear (by the text of the article) which they really mean. Raw data would help in this case.

*sigh*

Thanks, Gary!

Bon, I agree raw data is needed. However, the headline, the photo caption, and the sentence following the photo indicate to me that the reporter, at least, thought the conclusion applied to individuals, and was not simply a population conclusion.

“Imagine that half the people tested were experts and got it right every time, while the other half were dunces who got it wrong every time.”

Erm, I find it highly suspect that a dunce would get it wrong every time. That doesn’t really make sense. To get it wrong every time, you’d have to be 1. an expert who is intentionally trolling the experiment, or 2. a massochist who is drawn to bad wine, and the bad wine really does stand out.

The dunces would actually be likely to have about 50/50. So if the experts were anything above 50/50, the average would be above 50/50.

It’s like something an algebra teacher once remarked to me. One of us complained that our textbook was “always wrong”. He pointed out that if that’s the case, it’s easy, just always assume the opposite of what the book says. “The real problem” he said “is when the textbook is wrong half the time”. Again, in coding theory, a channel where 100% of the bits get switched by noise is just as good as a channel where 0% of the bits get switched by noise, and the worst possible corruption rate is 50% (anything else, even 50.001%, will allow arbitrarily low-noise communication using coding theory techniques)

Leave a Reply