From www.undeceivingourselves.org       (800 words 2 graphics)       Home       Fast-Find Index

Has wine a magnetic personality?
A repeat of our blind tests

By a panel of wine-loving skeptics doing their bit for Western Australia

In 2006 we tested the claim that magnets transform cheap plonk into a fuller, better tasting wine. Yes, the claim could revolutionise the WA wine industry, but we failed to confirm it, see our report Has wine a magnetic personality? Our blind tests say No on this website under Investigations by WA Skeptics > Strange Things. Four years later, undeterred by our depressing failure, we have again tested this claim with stronger magnets and an improved design using a similar red wine (a low-priced South Australian Shiraz) and ten testers.

Each tester could choose one or more pairs of numbered plastic cups. One cup contained about 20 ml of ordinary wine, the other contained the same amount of the same wine poured past magnets. Their order had been determined in advance using computer-generated random numbers. Could our testers, working alone or in pairs, tell which was which? Most testers managed 2-3 pairs of cups over 15-20 minutes, and the result was a total of 24 judgements. Not a huge number but about twice the average per wine of our previous tests, and much easier to analyse than the previous challenging combination of 4 wines x 4 magnet conditions.

Wine under test

Left: A total of 36 office magnets were affixed to the neck of the bottle by double-sided carpet tape, with north poles facing inwards on one side and outwards on the other to make sure the magnetic field passed through the wine. The field inside the neck was nearly 300 times stronger than the earth's magnetic field and about ten times stronger than the magnets we used in 2006. Next: Pouring the wine into clear plastic cups, which were deemed to be less offensive to oenophile sensitivities than the semi-opaque cups used previously, although they did make the inscribed numbers hard to read against the dark red wine. To make sure testing was blind, the pouring was done out of sight of the testers, and the pourers did not participate in the testing. Right: Wine testing in progress. Each tester had to decide which of the two samples had the superior flavour. The hypothesis was that the magnetised wine would be superior, so they had to pick one or the other -- don't knows or can't tells were not allowed.

As a control, the testers had to make the same judgements but without wine. They had to imagine taking two cups and deciding which one had the superior flavour. The result was 144 judgements, from which 120 sets of 24 judgements were sampled at random and compared with the 24 correct identities. Here 120 was chosen arbitrarily as a compromise between too few sets, say 10, and too many sets, say 1000, when compared with the 144 judgements being sampled. The result was a total of 120 comparisons, which in effect (much to the distress of true wine lovers but with conspicuous approval from our treasurer) provided 120 replications without using any wine. It also accommodated any response sets, where people (being human) like to answer A rather than B, or B rather than A. In this case the testers preferred B for 71% of the 24 actual judgements and 54% of the 144 imaginary judgements, both statistically significant at p<0.05, even though the occurrence of A and B in the correct identifiers was exactly the same.

Results of wine testing

Left: The form on which each tester recorded their judgements. Right: By a chi-squared test the distribution of imaginary judgements (red bars) shows no hint of being different from chance (black dots, X2 = 5.94, df = 10, p = 0.82). Of the judgements of actual wine, 14 out of 24 (58%) said the magnetised wine had the superior flavour. Is this statistically significant? No, see below.

In each test the probability of picking the magnetised wine purely by chance is 50%. In 24 tests the binomial significance level of getting 14 hits or more purely by chance is p = 0.27, which is a long way from being even marginally significant. Compared with the 16+4+5+4+1 = 30 imaginary replications that had the same score or more, the significance level is 30/120 = 0.25, or much the same. So although the results are in the right direction, they provide no significant support for the hypothesis that pouring cheap plonk past magnets improves its flavour. Our previous depressing failure has so far been confirmed.

Clearly, more research is needed with many more tests, many more testers, and (obviously) many more wines. The WA wine industry has never needed our help more than now. Wineries in other states are welcome to bid for our services. It will be tiring and demanding work but someone has to do it. Stay tuned!

From www.undeceivingourselves.org       (800 words 2 graphics)       Home       Fast-Find Index