<<

The Case of the Remembering Water

James Randi

FTER PARTICIPATING in the team's investigation of homeopathic claims asserted by the French laboratory INSERM AU200 in Clamart, France, I was led once more to the conclusion that self-deception and data flummery are not as rare in orthodox science as I had once suspected. Some 20 years ago, before I had looked into fields other than parapsychology, I was aware that a certain amount of impropriety took place in what is recognized as "regular" science. Those improprieties, I believed, were well publicized when they were discovered. Only with more experience did I learn that misconduct and even fraud are very often dis- covered but not reported by the academic world. Walter W. Stewart, who works for the National Institutes of Health in Bethesda, Maryland, spends much of his time looking into such matters. I met him in person for the first time during what has become known as the "Water with a Memory" episode. Stewart and I were asked to join , editor of Nature, to look into this most unusual claim. During the investigation, I recognized many of the earmarks of the experiments I had previously examined in the twilight zone of parapsychology. Many parapsychologists I have met are dedicated scientists, but often somewhat naive. They have, at best, peripherally significant data with which to support their beliefs. This kind of data was well described by John Beloff, once president of the Parapsychological Association, as "fragile." It does not stand up well under examination; nor are the experiments easily replicated. This is quite unlike "orthodox" scientific data, which are required to be robust in a variety of ways. Jacques Benveniste, director of INSERM U200, met us in his lab and prepared us for an introduction to Elizabeth Davenas, his chief experimenter. Now I have often said that there seems to be an automatic alarm system built into the back of my skull that sounds whenever I begin to receive suspicious information. That alarm began functioning. Said Benveniste: "Wait till you meet this woman. She is a person from another planet." I heard a faint tinkling. My colleagues squirmed a bit. "And," he continued, "she is a good church-going Christian." A distinct tinkling was heard this time. Benveniste went on: "Elizabeth is the only person in my lab who gets con- sistently good results with this test." The ringing was getting louder. "Her results are all recorded in notebooks with consecutively numbered pages." It sounded as if Big Ben were a block away, rather than across the English Channel.

142 THE SKEPTICAL INQUIRER, Vol. 13 That last comment was very significant. To me, it seemed to be offered as an assurance that the lab records had not been altered. Why would Benveniste specify such a fact? I suspected that perhaps—just perhaps—he was giving us a guarantee that his personnel would not and could not falsify data. That suspicion was reinforced when, later, he announced to us with his arms folded across his chest that he would not under any circumstances entertain the notion that his employees might be capable of "adjusting" any data, a possibility that we had not at that point even suggested to him. He told us that he did not want to hear the idea brought up. We were well warned. Our investigation revealed, first, that the mass of data generated from the Clamart lab could not have been obtained by the means described in the group's original Nature paper. That takes some explanation, which I now offer in a much simplified version. Any data obtained by counting or measur- ing samples (people, cells, weights, stars, etc.) is subject to the rule that it must conform to a "bell-shaped curve." This means that for any data-gathering system we can draw a curve on graph paper and plot each data-point against that curve. If we generate enough data, we will find that the data curve and the bell-shaped (Gaussian) curve agree closely. Any sufficiently large amount of data will follow this rule if the data have been properly obtained and recorded, since there must be errors in sampling or counting and there have to be extremes that are simply wrong but, because they are allowed for, do not affect the conclusions or the validity of the experiment. Stewart generated an appropriate Gaussian curve on his computer and fed the data obtained from Benveniste's lab books into it. The game was afoot: What we had before us was evidence that a great proportion of the lab data had been obtained by unorthodox methods. That is as kindly as I can put it. (See Figure 1.)

Benveniste data 1.2 (Actual distribution)

c 0.8 •o

a 0.4 Gaussian curve .o o (Expected distribution)

FIGURE 1

Winter 1989 143 We discovered that Benveniste had not done his experiments "blind," as he had described in his Nature paper. A blinded experiment (generally described) is one in which the experimenter does not know where the results should show up, for example, because there has been a randomization of samples. A nonblinded experiment is prone to being distorted—unconsciously or consciously—by anyone involved in it. Furthermore, we discovered that Benveniste's lab had omitted much negative data from their lab records because some mitigating circumstance had suddenly become apparent after it was discovered that a particular experiment gave negative results. For exam- ple, we were told that sometimes blood used in tests was, for a variety of reasons, found to be "unsuitable." Allergies, drug use, and fatigue on the part of the donor were a few of the reasons for throwing out the results of an experiment and failing to record them. Knowing what we did then, we concluded that the Nature paper had not described carefully several aspects of the research and was unacceptable. Though the three extensive experiments that we supervised at the 1NSERM U200 laboratory—except for being properly blinded—were done in exactly the same way Benveniste claimed to have done his experiments over the previous five years, the results were negative, a fact that Benveniste could not explain. But there was a move we then had to make. Stewart generated a second Gaussian curve on his computer and entered the new data obtained from our three experiments. Though this was a very small data base and the "fit" was therefore rough, the results were quite conclusive. (See Figure 2.) These data conform exactly as predicted by theory, while the much more extensive previous data in Figure 1 jump out at the observer as glaring evidence of something being very wrong. My personal opinion is that Benveniste believes in his findings. Yet we showed sound reasons, some of which I have described here, for the Ben-

FIGURE2

144 THE SKEPTICAL INQUIRER, Vol. 13 veniste claims to be rejected as evidence that water has a memory. The casual reader may be unfamiliar with the astounding flexibility of the minds of those involved in these remarkable claims. This elasticity was demonstrated during the most recent confrontation that Stewart and I had with Benveniste, on a London late-night television program. He actually announced that, after close examination of the results of the experiments we had supervised at his Clamart laboratory, he had found that, rather than damning his work, they had validated his theories! Once again, it boggles the mind. ... •

James Randi was a member of the Nature investigating team. He is a renowned conjuror and investigator and is a MacArthur Foundation Fellow.

Guide to the 'Dilution' Controversy

Here is a guide to the key reports on the remarkable case of Jacques Benveniste's claims and Natures subsequent investigation into the "dilution" controversy. They are arranged chronologically.—ED.

Primary Sources

Davenas, E.', et al. [including J. Benveniste]. "Human Basophil Degranulation Trig- gered by Very Dilute Antiserum Against IgE." Nature, 333:816-818, June 30, 1988. A one-paragraph "Editorial Reservation" by Nature appeared at the end of this paper. "When to Believe the Unbelievable" (Editorial). Nature, 333:787, June 30, 1988. Maddox, John, James Randi, and Walter W. Stewart. " 'High-Dilution' Experiments a Delusion." Nature, 334:287-290, July 28, 1988. Benveniste, Jacques. "Dr. Jacques Benveniste Replies." Nature, 334:291, July 28, 1988. "When to Publish Pseudo-science" (Editorial). Nature, 334:367, August 4, 1988. Benveniste, Jacques. "Benveniste on Nature Investigation" (Letter). Science, 241:1028, August 26, 1988. Maddox, John. "Maddox on the ''" (Letter). Science, 241:1585- 1586, September 23, 1988.

Notable Reports on the Controversy

Hudson, Richard L. "Nature Debunks Piece It Just Published That Supported Homeopath's Claims." Wall Street Journal, July 27, 1988, p. 30. Sullivan, Walter. "Report in Is Held to Be Flawed." New York Times, July 27, 1988. Dagani, Ron. "Nature Publishes Paper It Finds Unbelievable." Chemical & Engineer- ing News, July 4, 1988, p. 5. "Nature Sends in Ghostbusters to Solve Riddle of the ." New Scientist, July 21, 1988, p. 26.

Winter 1989 145 Rensberger, Boyce. "Research Called 'Shoddy' by Publisher of Article." Washington Post, July 27, 1988, p. A16. "Only the Smile Is Left." (Letters commenting on Davenas's original paper.) Nature, 334:375-376, August 4, 1988. Vines, Gail. "The Ghostbusters Report From ." New Scientist, August 4, 1988, pp. 30-31. Coles, Peter. "Benveniste Controversy Rages on in the French Press." Nature, 334:372, August 4, 1988. Pool, Robert. "More Squabbling Over Unbelievable Result." Science, 241:658, August 5, 1988. Langone, John. "The Water That Lost Its Memory." Time, August 8, 1988, p. 73. Wade, Nicholas. "Looking Hard at Science's Self-Scrutiny." New York Times, August 21, 1988. Dixon, Bernard. "Criticism Builds Over Nature Investigation" and "A Brief History of Dubious Science." The Scientist, 2(16):1, 4-5, September 5, 1988. Garfield, Eugene. "Contrary to Nature?" (Editorial). The Scientist, 2(16): 12, Septem- ber 5, 1988. "Unreproducible Results." Nature (Letters). 235:109, September 8, 1988. Five more letters. •

We will follow further developments in this remarkable episode with great interest. The case of the "Remembering Water" is not yet concluded. New information and perspectives are likely to emerge. The affair also raises important issues about the proper way to evaluate and treat highly controversial claims. Our pages will be open to future reports and discussions.—THE EDITOR.

Plan Ahead for the 1989 European CSICOP Conference

CSICOP's second European conference will be held at the resort of Bad Tolz, near Munich, Germany, from May 4 to 7, 1989. Our hosts will be the West German Society for the Scientific Investigation of Para-Science. Speakers will include European experts on topics of particular importance in Europe and members of the CSICOP Executive Council. Simultaneous translations into English, German, and Spanish will be provided. Programs and registration forms will be mailed to all European subscribers to the SKEPTICAL INQUIRER (including those in the U.K. and Ireland). If you live outside Europe and would like to attend, write to Mary Rose Hays, CSICOP, Box 229, Buffalo, NY 14215, USA, for a copy of the program and registration form.

146 THE SKEPTICAL INQUIRER, Vol. 13