The 'memory of water' — published with a warning, debunked by a magician

Jacques Benveniste was a serious immunologist: his INSERM laboratory in Paris was respected, and his team's earlier findings had appeared in Nature. In June 1988 the journal published his paper claiming that an antibody against IgE, diluted until the odds were against a single molecule remaining, still degranulated human basophils. Beyond Avogadro's number only water was left — so the water itself, he concluded, had to be remembering the antibody.

The claim was 'water memory', and it was exactly what homeopathy had always needed. Nature's editor John Maddox printed the paper with an unusual warning — there was no physical basis for such an activity, the journal said, and independent investigators would be sent to watch the experiments repeated. It was a bet: publish the impossible claim in the open, then settle it in public.

In July 1988 Maddox arrived at the Paris lab with two companions: Walter Stewart, an expert on scientific fraud, and James Randi, a professional magician. Over a week they watched the experiment run, then made it blind — labels off, no one knowing which sample was which. The effect vanished, scoring random. They found that the researcher at the bench, Elisabeth Davenas, examined tiny batches of data at a time and discarded results she judged insignificant. Nature published their verdict in an issue titled 'High-dilution experiments a delusion'.

Benveniste never conceded, calling the visit McCarthyite witch-hunting and repeating the experiments for the rest of his career. But the effect stayed gone: no independent lab has reproduced water memory under blind conditions since, and the episode became homeopathy's most-cited scientific citation — a paper in Nature that the same journal debunked in the same summer.