The Memory of Water Revisited

European homeopathic organizations have launched a campaign to convince policy makers that the proof for homeopathy has been definitively delivered, and that politicians should therefore take immediate action to recognize it as a part of official medicine. We take a look at this “proof” and conclude that it is, once again, a nonevent.

Homeopaths who try to gain recognition by the scientific community point their research in two directions: clinical and “basic” research. The first tries to provide evidence that homeopathic remedies have an effect on patients that is distinct from placebo. The second attempts to demonstrate that dilutions that contain no active ingredient still have a biological effect.

In 1988 newspapers in many countries carried frontpage articles announcing that the French scientist Jacques Benveniste had delivered the final proof for homeopathy. In an article in Nature,1 his team reported that homeopathic preparations, diluted far beyond Avogadro’s number and thus chemically free of any active component, still had biological activity. It was claimed that several independent researchers in other countries had reproduced these results. Since this discovery would upset basic principles of chemistry and physics, the editors had accepted this paper on the condition that they be allowed to witness a repetition of the test. The euphoria in the homeopathic community did not last very long. When the witnesses were present, the Benveniste team failed to reproduce the results under controlled conditions. This team of witnesses subsequently published an article in Nature calling Beneviste’s report a “delusion.”2

Other teams also tried without success to reproduce the results.3,4 For most scientists and some homeopaths, this was the end of an embarrassing affair, though some aficionados still maintain that Benveniste was the victim of conspiracy and censorship.5 Benveniste himself still believes and has even gone much further in his claim. He now calls it “digital biology” and claims to be able to record homeopathic “information” (or rays or vibrations?) on computer disks and send it by e-mail as an attached file. The receiver then simply has to play this information to a bottle of water and it becomes homeopathic medicine.6

Other researchers, including former collaborators of Benveniste and key figures from 2 French homeopathic industries, continued the efforts to demonstrate that such maximal dilutions have biological effects, and now claim to have definitive proof. Their articles were the subjects of a recent mailing campaign by the European Committee for Homeopathy. Letters such as the following were signed by the president or the research coordinator, and sent to politicians and ministries in July 2001:

Chers amis,

Il y a peu de temps, je vous signalais que les travaux sur «la mémoire de l’eau», si longtemps décriés, ont été confirmés par une étude multicentrique conjointe dans 4 universités européennes: UCL Belgique / Queen’s Univ. Belfast Irlande / Univ. Florence Italie / Univ. Utrecht Hollande / et 1 centre de recherche: Cerba France. Il n’y a plus aucun doute, les hautes dilutions homéopathiques provoquent bien des réactions biologiques! Vous trouverez en annexe les documents originaux . . .

Il est INDISPENSABLE que le monde politique comprenne qu’il est grand temps de tourner la page. Puisqu’aucun scientifique honnête ne peut encore considérer l’effet des dilutions homéopathiques comme un simple effet placebo, le politique se doit de faire bénéficier à toute la population européenne des bienfaits de l’homéopathie ou pour le moins de permettre d’accélérer le processus de reconnaissance scientifique de cette méthode de soins.

Nous attendons plus que des mots de sympathie. En toute amitié,

Co-ordinateur de la recherche ECH.


My dear friends,

A short time ago, I informed you that studies regarding “water memory,” long disparaged, have now been confirmed by a joint study of 4 European universities: UCL Belgium; Queen’s Univ. Belfast, Ireland; Univ. Florence, Italy; Univ Utrecht, The Netherlands; and 1 French research center: Cerba.

There is no more doubt: homeopathic dilutions do provoke biological reactions! You will find the original papers enclosed . . .

It is ESSENTIAL that governments understand that it is high time to turn over a new leaf.

Since no honest scientist can continue to say that the effects of homeopathic dilutions are mere placebo effects, politicians have a duty to help the whole European population benefit from the blessings of homeopathy, or at least to encourage the process of scientific recognition of this method of care and cure.

We deserve more than words of sympathy.

Fortunately, the senders had included copies of the 3 articles, since some of these publications are not to be found in most medical faculty libraries. The mention of 4 outstanding European universities made it worthwhile to have a closer look at this discovery, which, if found true, might lead to a Nobel Prize nomination. What do those trials tell us?


Belon P, Cumps J, Ennis M, Mannaioni PF, Sainte-Laudy J, Roberfroid M, Wiegant FAC. Inhibition of human basophil degranulation by successive histamine dilutions: results of a European multi-centre trial. Inflamm Res. 1999; 48(suppl. 1): S17–S18.

This article is only 1.5 pages long. The procedure is in vitro: basophils normally degranulate when mixed with anti-IgE serum. This is measured by coloring them, incubating them, and then counting them according to the degree of discoloration. Normally, the release of granules that contain histamine has a negative feedback effect on the further release of histamine. The homeopathic histamine solution—or rather, the solvent that contains the “memory of histamine”—is supposed to inhibit this discoloration (the dilutions are far beyond Avogadro’s number: from 1/1030 to 1/1038). The high dilutions are prepared in the new homeopathic way, i.e., not by shaking and “succussing,” but by high-speed vortexing. The reason this test has been chosen, it is explained, is that it is relatively simple to perform and does not require specialized equipment. One can also suppose that the similarity with the famous Benveniste experiment played a role. The tests were done in 4 different laboratories. A total of 3674 data points were collected, of which 840 were discarded as invalid. Several concentrations of antiIgE were tried and the concentration of 0.04 μg/ml was retained as the best for observation, presumably meaning that it gave the best results. At that concentration, 772 data points were retained as “valid.” In 3 out of 4 laboratories the results were claimed to be significant, so the total of all 4 is said to be significant. The table shows the results. The statistical comparisons were made using multiple analysis of variance (MANOVA). The authors conclude that the effect has been clearly demonstrated. They admit that the effect is small but insist that it is significant.



In the control group the degranulation varies between 45.8 and 51.6%. This is a rather broad range for reagents that were all prepared in the same laboratory and applied in the same standardized way, indicating that the standard is not especially stable and that chance plays an important role. The intervention data range between 35.7 and 47.4, so there is an overlap between this and the control group.

The 2 smaller centers (1 and 4) show differences of 15 and 17%. If my information is correct the biggest center (no. 2) was Belgian, so one could conclude that high dilutions do not work in Belgium.

Since I did not succeed in understanding the statistics used, I asked some biostatisticians their opinions and received the following comments:

  1. The use of MANOVA (multiple analysis of variance) does not make sense in this context. The correct procedure here should be an ANOVA [analysis of variance].
  2. To compare data specifically for each laboratory, a simple t-test for proportions suffices. Strangely enough, doing this computation for each laboratory gives statistically significant results only for lab 4, and even then barely so.Using the standard statistical formula for the test of two independent proportions (Wonnaccott & Wonnaccott Introductory Statistics), one obtains 95% confidence intervals of:
    • lab 1: [–.03; .21] ; not significant
    • lab 2: [–.06; .10]; not significant
    • lab 3: [–.05; .15]; not significant
    • lab 4: [.01; .23]; “barely significant”


  3. Even the significance as estimated above is widely overstated, since there were many (who knows how many?) comparisons and experiments.It is not clear what they mean by “number” on the column with this heading on table 1. I supposed their sample size, e.g., for lab 1 was 123 controls and 123 dilutions, but this is not clear. Does it mean, e.g., that for lab 1 there were 123 controls and 123 dilutions, or 60 controls and 63 dilutions, or what?
  4. I cannot make sense of many other details in their table, for instance why one of the values in the f-statistic “F” is missing (“n.a.”) (R.M. Varnier R. Almeida, written communication, August 29, 2001).

In the study the authors mention that 3674 measurements were made, but only 772 of these were retained. Many were rejected as “invalid data points” without any explanation. A letter to Prof. Ennis to ask for details about this and other statistical questions was never answered (J. W. Nienhuys, written communication, June 29, 2001). One wonders what the data were for the other anti-IgE concentrations. Why only the mention of those specific homeopathic dilutions? Did the authors try others? If so, why are those results not mentioned? All this points to vigorous fishing in the data.

The study is announced as a multicenter trial, but all participants received the same training, all reagents and materials were supplied from the same source, all the statistical analyses were done by the same person, and all data were “verified” by the French center. Thus it is not exactly a “multicenter” trial; it is more as if the work on one trial was divided among several centers. It is also interesting to note that the first author (Belon) is the son of the owner of a large French homeopathy factory.

Finally, this test ignores the homeopathic claim that shaking and diluting reverses the action of a product.

In conclusion, the suggestion that there is proof for the action of high dilutions is not substantiated by the results of this study.


Brown V, Ellis M. Flow cytometric analysis of basophil activation: inhibition by histamine at conventional and homeopathic concentrations. Inflamm Res. 2001; 50(suppl. 2): S47–48.

This study is essentially the same as the previous one, but includes more steps: marking the cells with a tracer that allows automated counting, which indirectly measures the amount of histamine production or degranulation: “the measuring was done by a 2-color (anti-IgE and anti CD63 R-PE) flow cytometric method to quantify the basophil activations and to investigate the effects of histamine on basophil activation.” Also, for unexplained reasons, some histamine solutions were first heated to 70°C and some samples were first frozen and then thawed. But the results of the freezing or heating cannot be found in the lone graph. It is only mentioned that heating neutralizes the effect, and freezing does not. The authors’ conclusion is that some homeopathic dilutions have a significant effect.


The whole text of the report is 1.5 pages long and rather enigmatic, but the graph of the results is remarkable: it shows the percentage of inhibition obtained by increasing dilutions from 2 to 40 zeros, though one wonders if it is possible to state the inhibition of degranulation as an exact value in view of the very indirect way of measuring. The first 3 blocks (102–106) show how the effect of histamine gradually diminishes with the dilution, but thereafter the results are erratic between +40 and –12%, meaning that some dilutions do close to nothing, some do even more than the real product, and some dilutions even have a negative effect! Does this mean that some cells were stimulated to degranulate even more? Since the authors write in the introduction that their study is aimed at the mechanism of the pathogenesis of asthma, that should mean that some dilutions (28 and 36 zeros) could provoke asthma. It would be interesting to see a repeat trial of this. The erratic results of high dilutions, however, are similar to Beneviste’s, and are reminiscent of the comment of Maddox et al. that Benveniste’s group had failed to consider contamination as an explanation for their findings. Intentional adulteration is another possible explanation.

Even if we assume that the results are credible (and that is all the more unlikely in view of the many steps, each with its own margin of error), we must conclude that the homeopathic effect appears and disappears. It even reverses at unpredictable moments: hardly a useful “remedy.”

There are some phrases in the study that make one wonder about the methodology; for example, “each measurement was performed at least 5 times.” Why not an exact number? One wonders if they kept measuring until they liked the results.


Wayne J, Lin Y, Tortella F. Neuroprotection from glutamate toxicity with ultra low dose glutamate. NeuroReport 2001; 12: 335–339.

The authors poison a culture of rat embryo neural cells with glutamate and then try to demonstrate that a homeopathic solution has a protective effect against this poisoning. The authors claim a 10% increase in survival rate.


Here, too, there are many intermediate steps that make the results untrustworthy, and the “effects” are very small indeed. Here, again, the results show unpredictable peaks. The authors used 2 kinds of cells: cortical and cerebellar. The cortical cells were not protected by the treatment except for a very small effect at dilutions 10–18 and 10–22. With cerebellar cells, the 10% effect was seen only at dilutions 10–20 and 10–30. These results are not more impressive than throwing dice.

The authors are at least honest enough to write that this study shows only that homeopathy could have a protective effect, but a very weak one, and only for some cells and only for minor damage. They also write that they cannot predict at what dilution there might be positive results. But in the same conclusion they stretch ethical boundaries by suggesting that homeopathy might help against Alzheimer’s and Parkinson’s diseases, a suggestion that is not at all warranted by this study.

This study was also financed by the Boiron homeopathic industry.


All 3 studies contain so many flaws that no conclusion about the in vitro effect of high dilutions can be substantiated. The studies are not about homeopathy as a system for the treatment of patients, they are not about treating diseases or symptoms, and they bear no relation to the diagnostic and treatment system of homeopathy. The only sense in which these studies are relevant to homeopathy is that they use the same extreme (“beyond Avogadro”) dilutions, containing not even a molecule of the original “active” substance, as those frequently used in homeopathic practice.

In conclusion, the results of these 3 studies do not justify the claims of European homeopaths in their letter to the politicians.


  1. Davenas E, Beauvais F, Amara J, et al. Human basophil degranulation triggered by very dilute antiserum against IgE. Nature. 1988; 333: 816–818.
  2. Maddox J, Randi J, Steward WW. High-dilution experiments a delusion. Nature. 1988; 334: 287–290.
  3. Ovelgonne JH, Bol AWJM, Hop WCJ, van Wijk RL. Mechanical agitation of very dilute antiserum against IgE has no effect on basophil staining properties. Experientia. 1992; 48: 504–508.
  4. Hirst SJ, Hayes MA, Burridge J, Pearce FL, Foreman JC. Human basophil degranulation is not triggered by very dilute antiserum against IgE. Nature. 1993; 366: 525–527.
  5. Fisher P. “The end of the Benveniste affair.” Br Homeopath J. 1999; 88(4). Available at: Accessed May 5, 2002.
  6. DigiBio Research Laboratorty Web site,