An issue of Ontario Skeptic contained a letter from Paul Greenwood, (“Science is open to radical, new ideas”) reporting on the “water memory” experiments of Dr. J. Benveniste, and offering the publication of these experiments as evidence of the willingness of the scientific community to examine new and unconventional ideas.

As most skeptics will realize, the claim that solutions can retain their effect when diluted many times, and indeed that the effect increases with dilution, is one of the fundamental tenets of the fringe medicine of homeopathy, but runs contrary to current knowledge in chemistry and biology. The results of this experiment, if validated, would therefore have lent credence to the claims of homeopaths.

However, additional information has come to light which forces us to reassess this research and its publication.

The initial report appeared in the June 30, 1988 issue of the British journal Nature, Vol. 333. Its results were sufficiently unusual that the editor of Nature began that issue with an editorial titled, “When to believe the unbelievable”. It is worth quoting extensively from that editorial:

Inexplicable observations are not always signs of the supernatural. That is what readers of the remarkable article on page 816 should keep in mind. They should also remember that Avogadro’s number, the number of molecules in a gram molecule of material is roughly UR1021, which naturally implies that most of the experiments with antibody solution reported by Professor Benveniste and his colleagues have been carried out in the literal absence of antibody molecules. For what the article shows is that it is possible to dilute an aqueous solution of an antibody virtually indefinitely without the solution losing its biological activity. Or rather, there is a surprising rhythmic fluctuation in the activity of the solution. At some dilutions, the activity falls off; on further dilution, it is restored.

There is no objective explanation of these observations. Nor is there much comfort for anybody in the explanation offered at the end of the article that antibody molecules once embodied in water leave their internal marks, as ghosts of a kind, on its molecular structure for there is no evidence of any other kind to suggest that such behaviour may be within the bounds of possibility….

Certainly there can be no justification, at this stage, for an attempt to use Benveniste’s conclusions for the malign purposes to which they might be put. There are some obvious dangers. In homeopathic medicine, for example, which works on the principle that very small concentrations of appropriate products may have consequences that far outweigh those expected of them, there will be a natural inclination to welcome Benveniste’s article as aid and comfort, but that would be premature, probably mistaken.

Nature also appended the following to the end of Benveniste’s article:

Editorial Reservation:

Readers of this article may share the incredulity of the many referees who have commented on several versions of it during the past several months. The essence of the result is that an aqueous solution of an antibody retains its ability to evoke a biological response even when diluted to such an extent that there is a negligible chance of there being a single molecule in any sample. There is no physical basis for such an activity. With the kind collaboration of Professor Benveniste, Nature has therefore arranged for independent investigators to observe repetitions of the experiments. A report of this investigation will appear shortly.

Apparently, the editors of Nature were sufficiently uncomfortable with Benveniste’s article that, after several revisions, they agreed to publish the article only if the researchers would permit their laboratory to be visited and their results investigated, by a team selected by Nature.

This team consisted of John Maddox, a journalist with a background in theoretical physics; Walter W. Stewart, a specialist in studies of errors and inconsistencies in the scientific literature and in the subject of misconduct in science, and James “The Amazing” Randi, whose name should be familiar to readers as a magician and skilled debunker of the paranormal and who was included on the team “in case the remarkable results reported had been produced by trickery”. After their investigation, the team published their findings in the July 28 issue of Nature, Vol. 334. They found in essence that the results were wholly erroneous.

Erroneous results found

Before their findings can be properly understood, it is necessary to understand the design of the original experiment. In the bloodstream is a variety of types of white blood cell, which perform various functions. This experiment focused on a particular type known as a basophil, so named because it contains granules which are selectively stained by basic (alkaline) dyes. Basophils carry on their surface a type of antibody called IgE (Immunoglobulin type E). If this antibody binds to a foreign substance for which it is specific, it causes the basophil to release its granules, which contain histamine and a variety of other substances responsible for the clinical signs of allergic reactions. The same result can be elicited by exposing the basophils to an antibody which binds to IgE, referred to as antiIgE.

In their experiment, Benveniste’s team began with a standard solution of anti IgE and repeatedly diluted it by 1:10. Each dilution was then added to a suspension of white cells, and the number of intact basophils was counted. (Basophils which have lost their granules do not pick up the specific stain and are therefore not readily distinguishable from certain other types of white cells.) This was done by placing a measured amount of the suspension in a counting chamber, a glass slide with a grid precisely etched on its surface, and counting the basophils under a microscope. The amount of degranulation was calculated by noting the difference in the number of basophils between test and control samples, expressed as a percentage.

Under the eyes of the investigative team, Denveniste’s researchers repeated their experiment seven times: three times using their normal procedure, once “blind” (that is, reading the dilutions in random order without knowing which ones they were), and three strict “double blind” experiments, in which no one present knew which samples corresponded to which dilutions until after they had been read. The “normal” runs produced the results that had been reported in the original paper. The fourth run produced very high peaks even at high dilutions, an effect comparable to the undiluted original sample. The three double blind runs produced the result that conventional science would have expected: a high initial peak, which drops off to a minimum when diluted.

Experiments don’t always “work”

The investigation team found the following flaws in the experiment:

1) The experiments do not always “work”. There are periods during which high dilutions produce the negative results that current science would expect, and these periods last for up to several months. These negative results were not reported.

2) The experimenters are far from unbiased. Two of the coauthors of the article receive salaries under a contract with a supplier of homeopathic medicines. This alone, of course, does not invalidate the experiment; most researchers expect to get some particular result when they run an experiment. It does, however, explain why the researchers, even now, refuse to admit that their results have been refuted. The investigators stated in their report that “the climate of the laboratory is inimical to an objective evaluation of the exceptional data”.

3) The peaks seen at high dilution, and described by Benveniste as “periodic” occur in fact at rather irregular intervals. Furthermore, Benveniste’s notebooks showed that the peaks do not occur at the same dilutions on successive runs. This alone suggests that the results reported are a random phenomenon, and not the positive result claimed.

4) There is a problem in experimental procedures of this type called “sampling error”. In essence, when a sample of blood is placed in the counting chamber, and the number of basophils is counted, the result will not be the same each time, but will vary randomly to either side of the true proportion. The branch of mathematics known as statistics provides formulas for calculating how much variation of this type can be expected for a given experimental setup. In the investigators’ report. they state:

At INSERM 200 [Benveniste’s lab], there seems to have gown up a less formal way of dealing with problems of this kind: When the reading of a diluted sample is greater than the control counts, the experimenter often counts the control sample again, on the grounds that the first reading “must have been wrong ….”

This procedure exaggerates to some extent the amount of basophil degranulation measured with reagents at high dilution. The practice makes the control values unreliable, and is a significant pointer to the laboratory’s disregard of statistical principles.

To someone with no training in statistics, this procedure might seem reasonable, or unimportant. In fact, it alone is sufficient to invalidate the results. Recounting only those control values which are “too low” has the effect of artificially inflating the control values, and thereby creating a purely artifactual difference between the control and sample values.

The investigators summarized their findings by stating, “We conclude that there is no substantial basis for the claim that antiIgE at high dilution (by factors as great as 10¹²⁰ retains its biological effectiveness, and that the hypothesis that water can be imprinted with the memory of past solutes is as unnecessary as it is fanciful.”

Dr. Scrimgeour is a veterinarian with Agriculture Canada’s Meat Hygiene Division and was a member of the executive of Ontario Skeptics when this article was written.