NATURE VOL. 334 28 JULY 1988
                                                  NEWS AND VIEWS                             ...
288                                                                                                             NATURE VOL...
NATURE VOL. 334 28 JULY 1988                                                                                              ...
290                                                                                                         NATURE VOL. 33...
NATURE VOL. 334 28 JULY 1988                                                                                              ...
Prochain SlideShare
Chargement dans…5

"High-dilution" experiments a delusion

1 800 vues

Publié le

Artículo de Nature donde "desmienten" un artículo que habían publicado, del Dr. Benveniste sobre la homeopatía.

Publié dans : Formation, Technologie
0 commentaire
0 j’aime
  • Soyez le premier à commenter

  • Soyez le premier à aimer ceci

Aucun téléchargement
Nombre de vues
1 800
Sur SlideShare
Issues des intégrations
Intégrations 0
Aucune incorporation

Aucune remarque pour cette diapositive

"High-dilution" experiments a delusion

  1. 1. NATURE VOL. 334 28 JULY 1988 NEWS AND VIEWS 287 "High-dilution" experiments a delusion The now celebrated report by Dr J. Benveniste and colleagues elsewhere is found, by a visiting Nature team, to be an insubstantial basis for the claims made for them. THE remarkable claims made in Nature tical physics. None of us has first-hand not to diminish the biological effective- (333,816; 1988) by Dr Jacques Benveniste experience in the field of work at ness of a molecule. and his associates are based chiefly on an INSERM 200. The experimental system has evolved extensive series of experiments which are We acknowledge that we might well from a test for assessing the susceptibility statistically ill-controlled, from which have found ourselves unable to get to grips of people to specific allergens. The guid-, no substantial effort has been made to 'with the work of the laboratory. But, on ing principle is that blood-borne allergens exclude systematic error, including the basis of our experience, we are confi- have the specific effect of interacting with observer bias, and whose interpretation dent that the design of the experiments the leukocytes known as basophils, caus- has been clouded by the exclusion of reported by INSERM 200 is inadequate as ing them to degranulate — that is, to measurements in conflict with the claim a basis for the claims made last month and release the contents of cytoplasmic that anti-IgE at "high dilution,, will de- that the defects we shall catalogue are a granules carrying histamine and other granulate basophils. The phenomenon sufficient explanation of the remarkable active substances provoking the symp- described is not reproducible in the results then reported. toms of asthma and hay-fever. ordinary meaning of that word. We believe that experimental data have These allergic reactions are apparently We conclude that there is no substantial been uncritically assessed and their imper- mediated at least in part by IgE molecules basis for the claim that anti-IgE at high fections inadequately reported. We attached to the surfaces of basophils (in dilution (by factors as great as 10120) believe that the laboratory has fostered the blood) or mast cells (in tissues). retains its biological effectiveness, and and then cherished a delusion about the Normally, degranulation is triggered by that the hypothesis that water can be interpretation of its data. the interaction of anchored IgE molecules imprinted with the memory of past solutes We are grateful to Dr Jacques with an antigen, but the same effect can be is as unnecessary as it is fanciful. Benveniste for his openness in discussing brought about by the use of anti-IgE — , We use the term "high dilution" reluc- most of the questions we raised with him. antibody prepared by injecting human tantly; these solutions contain no mol- He allowed us to borrow and to photocopy IgE into an animal of another species. ecules of anti-IgE, and so are not solutions the relevant laboratory notebooks, which (INSERM 200 uses goat anti-IgE at a con- in the ordinarv sense. "Solute-free solu- were invaluable for our investigation. We centration of 1 mg cm-3 sold by the Dutch tion" would similarly be illogical. have every reason to believe that Dr company Nordic.) Our conclusion is based on a week-long Benveniste was (and, perhaps, still is) The laboratory notebooks provide visit to Dr Benveniste,s laboratory, the convinced of the reality of the phenomena ample evidence that this expected INSERM unit for immunopharmacology reported in his article. We are also in the degranulation is a maximum between and allergy (otherwise INSERM 200) at debt of several of Dr Benveniste's col- log(dilution) 2 and 4. Clamart, in the western suburbs of Paris, leagues, especially to Dr Elisabeth Benveniste described the published during the week beginning 4 July. Among Davenas. On her fell most of the burden procedure as a "simple experiment",. A other things, we were dismayed to learn of demonstrating the standard dilution buffered solution of anti-IgE is serially that the salaries of two of Dr Benveniste's experiments and of repeating them in a diluted by a factor of 10 by transferring coauthors of the published article are paid blinded protocol under our scrutiny. We measured volumes from one test-tube to for under a contract between INSERM know that our report will be a disappoint- another. Pipette tips are discarded after 200 and the French company Boiron et ment to the laboratory. We are sorry. each transfer. Measured volumes of re- Cie., a supplier of pharmaceuticals and What follows is a narrative account of suspended white cells derived from homoeopathic medicines, as were our our visit and a summary of our conclu- human blood are transferred to wells in a hotel bills. sions. polystyrene plate. To each of these is Benveniste's results are being widely Our investigations concentrated added a measured volume of serially interpreted as support for homoeopathic exclusively on the experimental system on diluted anti-IgE or buffer as a control. The medicine. In the light of our investigation, which the publication was based. During wells are incubated for 30 minutes at we believe that such use amounts to our week in Paris, we resisted several 37°C. An acidic solution of toluidine blue, misuse. proffered opportunities to examine other which stains intact but not degranulated Our visit and investigation were systems in which high dilution is claimed basophils red, is added and the numbers of preconditions for the publication of the recognizable basophils counted on a original article. We acknowledge that we haemocytometer slide. Anti-IgG, which are an oddly constituted group. One of us does not degranulate basophils, is used as (J.R.) is a professional magician (and also a control. a MacArthur Foundation fellow) whose We were surprised to learn that the presence was originally thought desirable experiments do not always "work,,. There in case the remarkable results reported have been periods of several months at a had been produced by trickery. Another time during which solutions at high dilu- of us (W.W.S.) has been chiefly concern- tion have not degranulated basophils. ed, during the past decade, in studies of Indeed, the laboratory had just emerged errors and inconsistencies in the scientific from such a period. (Speculation at the literature and with the subject of laboratory is that the distilled water may misconduct in science. The third (J.M.) is Fig. 1 A demonstration degranulation, the have been contaminated, or otherwise a journalist with a background in theore- first of the three open experiments. made unsuitable.) It also appears that
  2. 2. 288 NATURE VOL. 334 28 JULY 1988 NEWSANDVIEWS- veniste volunteered that "we've not seen we should eagerly have accepted Ben- one like this before". The odd feature of veniste's invitation to devise a blind the curve is that the activity of the diluted experiment. We set out to devise a pro- anti-IgE is, at its peak, identical with that cedure that would be watertight. We of anti-IgE at log(dilution) 3 — presum- asked that three samples of blood should ably the point at which the natural be run. The serial dilutions would be pre- degranulating effect of anti-IgE is a pared by Dr Davenas, secretly coded by us maximum. before being transferred to wells for incu- We raised with Dr Benveniste and his bation and staining by her. colleagues the obviously relevant question In a small laboratory, procedures like of the sampling error. We were astonished this are inevitably and understandably dis- to learn, in the discussion of our conclu- ruptive. At INSERM 200, the sense of sions at the end of our visit, that neither melodrama was further heightened by the Dr Benveniste nor his colleagues seemed general recognition of the importance of to be aware of what sampling errors are. the trial, and by the precautions necessary We provided a simple explanation, com- to ensure that the code would not be plete with an account of what happens known to others than ourselves as well as when one pulls a handful of differently by the need that the one of us with a Fig 2 The fourth demonstration experiment coloured balls from a bag, to argue that reputation for sleight of hand (J.R.) could (read "blind") with unexpectedly high the sampling error of any counting i be shown to have been kept away from the peaks (see text). measurement must be of the order of the test-tubes containing the serial dilutions. bloods that "do not degranulate" are often square root of the number to be counted. This was done by arranging that encountered; we were informed that, in On several occasions, Benveniste called Davenas should carry the diluted anti-IgE this event, data are recorded but not these "theoretical objections". solutions in stoppered test-tubes to a included in analyses prepared for publica- Ironically, he is himself one of the three separate room, where their contents tion. Even so, the source of blood for the authors of a paper published in 1981, in would have been transferred to previously experiments is not controlled, except that which just this issue had been addressed in labelled tubes as determined by counters an attempt is made not to use blood from a superficially similar situation (Petoit, drawn at random from a bag. The coding people with an allergy. J.F., Sainte-Laudy, J. & Benveniste, J. procedure was monitored by a video We witnessed a total of seven runs of Ann. Biol. clin. 39, 355; 1981), and which camera operated by Randi, who was this experiment, of which three were appears to be the justification of the dot- thereby prevented from touching any- routine repetitions of the standard pro- ted line drawn at about 20 per cent (cor- thing else. (We have a record of the pro- cedure. For the fourth experiment, responding to two standard deviations) on ceedings on an unbroken reel of tape.) samples of diluted IgE were transferred by the per cent degranulations of intact We made two last-minute changes in one of us (W.W.S.) to wells in a plastic basophils after axis. the planned procedure. First, we included plate in a random sequence and then read That brief paper deals exclusively with 5 control tubes containing only buffer. blind by Dr Davenas. All four of these the effect of sampling errors (not other experiments, the last after decoding, gave kinds of errors) on the interpretation of results described as positive by Ben- measurements of intact basophils after veniste. But three further sequences of white-cell suspensions had been allowed counts of stained basophils in three to react with allergens via their attached further strictly blind experiments gave IgE molecules. Even now, at the Clamart negative results (see below). laboratory, provision is made for the Figure 1 shows results gathered in the measurement of two control samples. first group of experiments. The ordinate is Among other things, the paper provides a the decrease (compared with the control) statistical test for telling when the differ- of the numbers of stained basophils at ence between the two control values is dilutions ranging from 10-2 to l0-30. In each statistically significant at the 5 per cent case, the left-most peak is that expected level, in which case people using the pro- from the interaction between anti-IgE and cedure as a diagnostic test of allergy are IgE bound to basophils. The number of advised to start their experiment all over stained basophils increases to near its con- again. trol value at log(dilution) of between 5 At INSERM 200, there seems to have and 7 (0 per cent degranulation, called grown up a less formal way of dealing with "achromasie"); the unexpected phenom- problems of this kind; when the reading of enon is that the graphs then reach a series a diluted sample is greater than the control of three or four further peaks with increas- counts, the experimenter often counts the ing dilution. control sample again, on the grounds that These are the successive peaks of the first reading "must have been wrong". activity said in the original article to occur This happened when Dr Davenas was in a periodic fashion, and whose position counting the first of the first group of was said to be reproducible. It is clear experiments. from the four graphs that this claim is not This procedure exaggerates to some Fig. 3 Records for the first two blind obviously supported by this data. The extent the amount of basophil degranu- experiments (5-7 inclusive), showing laboratory notebooks confirm that the lation measured with reagents at high sampling noise only below the expected position of the peaks varies from one dilution. The practice makes the control decline of degranulation with increasing experiment to another. values unreliable, and is a significant dilution. Note that the ordinate extends The data in the fourth experiment pointer to the laboratory's disregard of below zero on the degranulation scale (to appear different from those recorded statistical principles. accommodate sampling errors above as earlier in the laboratory. Indeed, Ben- In these circumstances, it is natural that well as below the control values).
  3. 3. NATURE VOL. 334 28 JULY 1988 289 NEWS AND VIEWS pulled away, but inspection of the alumin- ium foil allowed him to pronounce himself satisfied that the code had not been read. Then came the decoding — one person singing out numbers to another. So do the numbers make sense? Six numbers into the record of the first plate to be read, Benveniste said "that patient isn't degranulating, try another,,. So we did — first the parallel readings by Dr Beauvais, then the remaining two experiments. In the event, the results of all three experiments were similar. The Fig. 4 Comparison of measured departures of duplicate normalized readings from their anti-IgE at conventional dilutions caused means with the gaussian distribution expected. degranulation, but at "high dilution" there was no effect. Blood from three Second, having been warned that homoe- results of the double-blind experiments sources in a row degranulated at ordinary opathists might regard the data as invalid might be unreliable because the observers dilutions but not at homoeopathic dilu- if solutions were decanted from one set of had been exhausted by our demands. tions. Each of the three experiments was a tubes to another, we removed the num- Others working in this field recognize failure. bers written with a felt pen on the original the difficulty of counting basophils tubes, replacing them with numbered (roughly 1 in 100 among leukocytes), pre- Conclusions labels which Randi assured us were ferring instead to measure the histamine We conclude that the claims made by tamper-proof. The code itself was even- released on degranulation. This practice is Davenas et al. are not to be believed. Our tually folded in aluminium foil, enclosed not followed at INSERM 200 because, we conclusion, not based solely on the cir- in an envelope specially sealed by Randi were told, of previous failure to record cumstance that the only strictly double- and then taped to the laboratory ceiling histamine release (as distinct from the dis- blind experiments we had witnessed for the duration of the experiment. appearance of stained basophils) at high proved to be failures, may be summarized We also arranged a second step of dilution (whence the term "achromasie"). as follows: coding just before the slides were We began to break the codes by lunch- • The care with which the experiments counted. One of (W.W.S.) took responsi- time on our last day, the Friday. When the reported have been carried out does not bility for pipetting, after securing the slides had been matched to the wells from match the extraordinary character of the agreement of Davenas and Benveniste which their samples had derived, but claims made in their interpretation. What that his technique was satisfactory. Both before the appropriate dilutions had been we found, at Clamart, was a laboratory the laboratory procedures and the codes assigned to them, there was a great sense procedure possibly suitable for the appli- themselves were recorded on video tape. of light-heartedness in the laboratory, no cation of a well-tested bioassay, but un- The plates containing the stained cell sus- doubt at the prospect that the ordeal suitable as a basis for claiming that anti- pensions were stored in a box (sealed by would soon be at an end. Benveniste, IgE retains its biological activity even at a Randi) in a cold room until read, in ran- glancing at the half-decoded data, even log(dilution) of 120. In circumstances in dom order. The second plate took longer offered to predict where the peaks and which the avoidance of contamination to read, partly because each well was read troughs would fall in the data. His offer would seem crucial, no thought seemed to in duplicate by each observer (Dr was accepted. But his predictions proved have been given to the possibility of con- Davenas and her colleague, DR Francis to be entirely wrong. tamination by misplaced test-tube stop- Beauvais), partly because the cells of the We asked at this stage for criticisms of pers, the contamination of unintended second plate were only faintly stained and the conduct of the trials, but were given wells during the pipetting process and were thus difficult to read. none. To the question what would be said general laboratory contamination (the Whatever the three runs would provide, if the two observers had recorded experiments we witnessed were carried we were especially anxious to derive some degranulation peaks, but at different high- out at an open bench). We have no idea objective estimate of the intra- and inter- dilution values, Benveniste said that what would be the effect on basophil observer measurement errors. We had would still constitute success. degranulation of the organic solvents and been told at the outset, by Benveniste, Opening sealed envelopes is Randi's adhesives backing the scotch tape used to that Dr Davenas was not merely excep- expertise. He found that the sealed flap of seal the polystyrene wells overnight, but tionally devoted to her work but the one in the envelope had detached itself at a sur- neither does the laboratory. whose hands the experiment most often prisingly straight angle when the scotch The design of the experiments hardly "works". He said that she usually "counts tape attaching the code to the ceiling was matches the nature of their interpretation. more cells" than other people. Dr Beauvais, who was also said to be excep- tionally skilled, read the slides separately from Dr Davenas, but at the same time. On this occasion, the sampling errors missing from most of the laboratory records did indeed appear. The duplicate measurements in our strictly blinded experiments were especi- ally important. First, they show that sampling errors do indeed exist, and are not "theoretical objections". Second, they show that the two observers were counting as accurately as could be expected, which Fig. 5 Same as Fig. 4 except that data derive from duplicated readings within the blind gives the lie to the later complaint that the experiments only.
  4. 4. 290 NATURE VOL. 334 28 JULY 1988 NEWS AND VIEWS For example, one would have thought that tested positive in this trial revealed not counting wells at least in duplicate would merely immunoglobins but other protein have been an elementary precaution contaminants apparently identical with against gross errors. The second of our materials in the original IgE vial. One of strictly blinded experiments seems to be the participants (Professor Meir Shinitsky one of the few in which something of this of the Weizmann Institute) then withdrew kind had been attempted. as a putative co-author. The laboratory seems to have been Since then, there have been two de- curiously uncritical of the reasons why its velopments in Israel — a series of experi- experiments do not, on many occasions, ments carried out independently of "work". For example, we were told that Benveniste,s laboratory and a further the best results were obtained when cells blinded experiment. Data from the latter were left in the cold-room overnight are unfortunately not available. Maitre before counting, but there has been no Fig. 6 Two duplicate Italian runs showing Simart, a legal official at Clamart who investigation of that phenomenon, or of high degranulation, but discordantly. held the codes, is said not to have had time the reports that taking a second sample shared belief at Clamart in the reality of to decode them. from a single well gives odd results (an the phenomenon reported last month, These measurements are nevertheless, effect not apparent in our double-blind and its potential importance, it is mysti- to judge from the documents we have experiments). fying that duplicate and blind counting is seen, stronger evidence than any we found • The phenomena described are not not routine. at Dr Benveniste,s laboratory to support reproducible, but there has been no serious • The climate of the laboratory is inimical his claims. But we do not have the infor- investigation of the reasons. We have to an objective evaluation of the excep- mation to evaluate them. referred to the fact that some blood yields tional data. So much is readily apparent negative results, and that there are from the way in which experiments are Postscript periods of time when no experiments described as successes and failures, by the We presented the substance of these work. But the laboratory notebooks show use of the word ''working", to describe conclusions to Dr Benveniste and his col- great variability in the positions at which experiments yielding a positive result, and leagues immediately after the strictly peaks occur. by the several speculations we were blinded experiments were decoded. The • The data lack errors of the magnitude offered, without experimental evidence in discussion that followed was inevitably that would be expected, and which are their support, to explain the several tense. Benveniste acknowledged that his unavoidable. This is best illustrated by failures the laboratory has experienced. experimental design may not have been Fig. 3, whose two graphs have been con- The folklore of high-dilution work "perfect", but insisted (not for the first structed from data recorded by Dr pervades the laboratory, as epitomized by time) that the quality of his data was no Davenas from samples supposedly identi- i the suggestion that decanting diluted worse than that of many papers published cal with each other, usually measurements solution from one tube to another might in Nature and other such journals. of control samples but also including some spoil the effect and the report that the One of us (J.M.) said it would be best if duplicate runs. The recorded values have repeated serial dilution by factors of three Benveniste would withdraw the published been normalized by subtracting the mean and seven (rather than ten) always yields article, or at least write to Nature to qual- and dividing by the square root of the negative results. ify his findings and their interpretation, in mean (the expected sampling error). If the which case we would not publish this only source of error were sampling error, Collaborations report. It was mutually agreed that noth- the standard deviation of the plotted curve We have not been able to pay as much ing would be said publicly until 28 July. should be unity (1). Other sources of attention as we would have wished to the But Benveniste said that the laboratory error, for example, experimental variabil- data collected at other laboratories and would work through the weekend "and all ity, could only increase the standard cited in Davenas et al., but we have ex- next week" to prove the reality of the deviation. But Fig. 3 shows that repeat amined documentary evidence available. pheonomenon. observations agree more closely than Supporting data were said to have come Our greatest surprise (and disappoint- would be expected from the underlying from Rehovot (Israel), Milan and ment) is that INSERM 200 seems not to distribution. This is a well-known effect Toronto. Dr Benveniste told us we could have appreciated that its sensational that sometimes affects duplicate readings not see the Toronto data, described as claims could be sustained only by data of by the same individual, but the magnitide preliminary, without the consent of the exceptional quality. Randi put the point of the effect in this case calls into question authors, who could not be telephoned. best, during our Friday discussion, by say- the validity of the readings. This artefact is The data gathered in Israel and Milan ing: "Look, if I told you that I keep a goat nevertheless not apparent in the blinded are, apparently, significant. Figure 6 is in the backyard of my house in Florida, duplicated readings. typical of the data from Milan. Though and if you happened to have a man near- • No serious attempt has been made to there are no duplicate measurements and by, you might ask him to look over my eliminate systematic errors, including therefore no direct evidence of sampling garden fence, when he'd say 'That man observer bias. It is true that the laboratory error, there is also some evidence of keeps a goat'. But what would you do if I notebooks record experiments in which degranulation at high dilution. Without said, 'I keep a unicorn in my backyard'? anti-IgG has been used as a control; we knowing more about the circumstances, We have no way of knowing whether the were surprised to find that the IgG control we are unable to comment. point was taken. run reported by Davenas et al. (their The Israeli data are more extensive. Eventually, there was no more to say. Fig. 1b) was carried out at a different time The first trials were in March 1987, during We shook hands all round, sped past the from the run with IgE published in the a visit to Rehovot by Dr Davenas. The common-room filled with champagne same figure. most remarkable of several successful bottles destined now not to be opened and Most of the data recorded in the trials was her correct identification of into the lens of a news agency photo- laboratory notebooks derive from seven high-dilution tubes out of ten grapher summoned for the happier event. experiments in which the same person has presented to her blind. Even so, the report John Maddox been responsible for the sequential dilu- (to Benveniste) of the trials was cautious, James Randi tion, plating out and counting. Given the Later, analysis of the tubes which had Walter W. Stewart
  5. 5. NATURE VOL. 334 28 JULY 1988 291 NEWS AND ViEWS- Dr Jacques Benveniste replies: open (7 series. 183 samples) control wells. Did the "experts" understand that the real controls are water or anti-IgG most AMAZINGLY, J. Maddox, with all his and knew both codes (dilution and often paired with anti-IgE (Fig. 16)? They experience, fell with us into the trap set by counts). analysed a few curves out of 1,500 pages, a squad of "self-appointed keepers of the Here is another hard-to-believe but most positive data are anyhow way off scientific conscience", "with no substan- incident: Stewart imposed a deadly silence 1 or 2 standard deviations. Other allergy tial scientific published record" (J. in the counting room, yet loud laughter tests correlate with degranulation (refer- Maddox, Nature 333, 795; 1988). Their was heard where he was filling chambers. ence in article), so why is it that our statis- amateurism, the climate they created in There, during this critical process, was tics fit for 40 to 70 per cent degranulation the five days of our ordeal, their inability Randi playing tricks, distracting the tech- at regular ligand concentration and not for to get to grips with our biological system nician in charge of its supervision! the same at high dilution? and their judgement based on one dilution It will now be clear what a mockery of Similar double-blind experiments (Br. series dismiss this inquiry altogether. scientific inquiry this was. Only the con- J. clin. Pharm.) were under the control of Who, with even the slightest research stant implication that we had something to an INSERM statistician, using a better background, would blot out five years of hide (the squad left with 1,500 photo- non-parametric test seemingly unknown our work and that of five other labora- copies!) prevented me from stopping this to our visitors. Then, the report auto des- tories on such grounds? masquerade. On one blood, basophils troys the statistical bias, declaring it "not For two years, I asked Nature to check could barely be counted. On the two applicable to all " (how many?) "data, for our data. But the magician and the invigi- others, controls ranged from 40 to 81 for example in the 4th experiment" similar to lator defined above worried me deeply. one operator, from 35 to 61 for the other, Fig. 1b or to the double-blind tests super- Mr Maddox assured me that he would the worst ever. vised by our Dean and a bailiff or Israeli prevent any wrongdoing. In fact, a tor- Duplicates such as 39-63 were found; if scientists (tables). nado of intense and constant suspicion, 39 were right, degranulation would have Being statistically sound (which is fear and psychological and intellectual been 61 per cent at dilution 22. Thus, the "bloody obvious" using common sense), pressure unfit for scientific work swept first three open and blind tests worked, are all these results "made up" as snapped our lab. Furthermore, these lesson-givers controls being impeccable, whereas on the at me by Stewart, the very referee who were astonishingly incompetent. In spite last days the test worked poorly mainly cleared the paper with raw data and statis- of my demands, no programme was set due to erratic controls. Something hap- tics in hand? Why then accept a paper on beforehand. pened, probably the work load and modi- 13 June to publish June 30th to destroy on There were performed in 5 days 3 x 30 fications enforced by the "expert". 8 July data so easily spotted as wrong or ten-fold dilutions, preparation and All in all,the judgement is based on one made up? Is it a display to the world of the degranulation (35 tubes each) of 7 leuko- dilution tested on two bloods in awful almighty anti-fraud and heterodoxy cyte samples, and eye scan of 300 cham- technical and psychological conditions. squad? Lip service is paid to our honesty; bers (about 20,000 basophils). Half of that Outrageous! Then, the team flew away in yet accusation of cheating was rampant, as is way beyond the weekly individual limit. minutes, not leaving behind any report, shown by dismissal of the 4th experiment, The first two days of the week were spent nor even the data that I had to collect at Randi's mere presence and his lengthy on four open experiments. The first blood Stewart's hotel that night! The report is examination of the supposedly violated did not react even to high anti-IgE, but the filled with inaccuracies and distortions. code. This impinges on our honesty and three other results were superb. The Just a few: does the fact that homoeopathic scientific ability but also, without fourth (counted blind upon our insistence) companies are paying two researchers examination, on the other participating was "incredible": 70-75 per cent degranu- (contract approved by INSERM adminis- laboratories, which is unacceptable. I lation at dilutions 10, 16/18, 22, similar to tration) mean that they order them into welcome academic exchanges on errors, if Fig. 1b of the article, controls varying by improper conduct? How about research in any, but will no more stand suspecting us the usual 15. — or supported by — industry, including or our associates. Then Stewart, with his typical know-it- numerous Nobel prizes? We could not More, I now believe this kind of inquiry all attitude, called these results, blind self-finance a long-term international must immediately be stopped throughout though they were, valueless; that implies cooperation nor the expenses of this large the world. Salem witchhunts or fraud before counting. The third day, a group of investigators. Did the source of McCarthy-like prosecutions will kill new dilution series was single-coded in the money influence their judgement? science. Science flourishes only in free- front of a video camera, involving two What a level of argument! dom. We must not let, at any price, fear, major professional errors since all the visi- The Scotch tape was placed above all blackmail, anonymous accusation, libel tors knew the code, when "to believe the wells (see controls below). Repeating a and deceit nest in our labs. Our colleagues unbelievable"? (The witness camera wrong count? This detects counting are overwhelmingly utmost decent could not record time, nullifying that part errors, especially when stressed by people, not criminals. To them, I say: of the procedure.) The code, wrapped pointed microphones and camera. The never, but never, let anything like this into aluminium foil and in an envelope, two closest counts are chosen at risk of happen — never let these people get in was taped to the ceiling! being all wrong, as one recalibrates an ill- your lab. The only way definitively to The next day, the hysteria was such that tuned machine, even in a blind experi- establish conflicting results is to reproduce Maddox and I had to ask Stewart not to ment. The central argument bears on them. It may be that all of us are wrong in scream. He had decided also to blind the sampling errors and statistics of which we good faith. This is no crime but science as counting (an overkill) and to fill the are so aware that we performed numerous usual and only the future knows. D chambers, using a modified untested control experiments. They show similar standard deviations and variances in 24/28 J. Benveniste is at INSERM U200, Universite method (two other serious errors). Refer- Paris-Sud, 32 rue des Camel, 92140, Clamart, ees must respect experimental design and comparisons of blind (4 series, 90 samples, France. not take part in it. This one was untrained without the Israeli experiments) versus