Content uploaded by Francis Beauvais
Author content
All content in this area was uploaded by Francis Beauvais on Nov 06, 2017
Content may be subject to copyright.
E
MERGENCE OF A
S
IGNAL FROM
B
ACKGROUND
N
OISE IN THE
“M
EMORY OF
W
ATER
”
E
XPERIMENTS
:H
OW TO
E
XPLAIN
I
T
?
Francis Beauvais, MD, PhD
1#
After more than 20 years, the case of the “memory of water”
still has not been resolved satisfactorily. After the affair with the
journal Nature, Benveniste extended his results on high dilutions
to an “electromagnetic biology” and then to a “digital biology,”
where electromagnetic signals supposed to be emitted from bi-
ologically active solutions were said to be stored on magnetic
memories. Although the results obtained by Benveniste and
coworkers were obvious, the difficulties in reproducibility by
other teams created doubt of the reality of the alleged phenom-
enon. In a first step, we analyzed a set of experiments obtained
by Benveniste’s team in the 1990s. We quantified the relation-
ship between “expected” effects (ie, labels of the tested samples)
and apparatus outcomes, and we defined the experimental con-
ditions to observe significant correlations. We concluded that
the results of these experiments were related to experimenter-
dependent correlations, which did not support the initial “mem-
ory of water” hypothesis. The fact that a signal emerged from
background noise, however, remained puzzling. Therefore, in a
second step, we described Benveniste’s experiments according to
the relational interpretation of quantum physics of C. Rovelli. In
this interpretation, the state of a system is observer-dependent
and the collapse of the wave function appears only in the states
relative to a given observer. This interpretation allowed us to
elaborate a model describing Benveniste’s experiments in which
the emergence of a signal from background noise was described
by the entanglement of the experimenter with the observed sys-
tem. In conclusion, the pursuit of the experimental “proof” to
support the “memory of water” hypothesis has prevented other
interpretations. Although our hypothesis does not definitely dis-
miss the possibility of “memory of water,” the experimenter-
dependent entanglement could be an attractive alternative inter-
pretation of Benveniste’s experiments.
Key words: Memory of water, high dilutions, digital biology,
scientific controversy, experimenter effect
(Explore 2012; 8:185-196. © 2012 Elsevier Inc. All rights reserved.)
“My own conviction is that it remains to be shown that
there is a phenomenon to be explained”
John Maddox (1988)
1
INTRODUCTION
In June 1988, an article published in the prominent journal
Nature was the start of a scientific controversy that remains
unresolved today.
2
The experiments described by this article
suggested that serially diluted solutions of antibodies, shaken
between each dilution, retained their specific biological activity
to stimulate white blood cells called basophils. For any reader, it
was a shock to realize that less than one active molecule was
present in the highly diluted solutions. If true, these results were
a ground-breaking discovery. But how to explain them if no
active molecule from the initial solution was present? Were
“ghosts” of molecules produced by the dilution process, or did
the molecules “imprint” a template in water?
The article was accompanied with an unusual editorial reser-
vation, which warned the reader of accepting these odd results.
Another unusual circumstance was that the publication of the
manuscript had been made on the condition that Nature’s team
could perform an inquiry into the laboratory of Jacques Ben-
veniste, the lead author of the article. Performing the inquiry
after—and not before—the publication was nevertheless a surpris-
ing method.
This publication was the result of a long battle of Benveniste
with the editorial team of Nature, more particularly with the
Editor John Maddox, who feared that the proponents of home-
opathy would find scientific support with the paper once it was
published. Indeed, although the word homeopathy was not pres-
ent in the manuscript, high dilutions are one of the principles of
this disputed alternative medicine, and it was clear that the ex-
periments had been performed in the context of industrial con-
tracts with homeopathy firms.
However, this manuscript could not be easily discarded by the
editorial board of Nature because Benveniste was a reputed se-
nior director of INSERM, the French medical research organi-
zation. He was one of the discoverers in the 1970s of the platelet-
activating factor, which belongs to a new class of inflammatory
mediators. In other words, Benveniste was not an isolated eccen-
tric scientist but a member of the scientific establishment. More-
over, Benveniste was a tenacious man and during the two years
of negotiations with Nature, he answered all the demands—
which were perhaps also designed to postpone the final deci-
sion—and particularly the request for the reproduction of the
results by other laboratories.
One month after the publication of the article, the five-day
inquiry began in the laboratory of Benveniste at Clamart, in the
suburbs of Paris. To compose the team of investigators, Maddox,
himself a former theoretical physicist, did not recruit specific
1 91, Grande Rue, Sèvres, France
#Corresponding Author. Address: 91, Grande Rue, 92310 Sèvres, France.
e-mail: beauvais@netcourrier.com
185
© 2012 Elsevier Inc. All rights reserved EXPLORE May/June 2012, Vol. 8, No. 3
ISSN 1550-8307/$36.00 doi:10.1016/j.explore.2012.02.004
HYPOTHESIS
experts but the pseudoscience debunker and stage magician
James Randi and the fraud investigator Walter Stewart. Indeed,
Maddox, as he reported later, suspected that someone in the
entourage of Benveniste was cheating. During the first three
days, the investigators examined the laboratory notebooks and
observed one scientist performing high-dilutions experiments.
Rapidly, the suspicion of cheating was abandoned; nevertheless,
the results of the ongoing experiments, including one blind
experiment, were in favor of an effect of high dilutions. From
this moment, the investigators involved themselves in the exper-
iments, not only in the blinding of the highly diluted samples,
but also in the pipetting of the cell suspensions for their alloca-
tion to the two members of Benveniste’s team who counted the
stained basophils under a microscope. The experiments per-
formed during these last two days with the assistance of the
investigators failed. All the experiments of the inquiry week have
been described in details.
3
The report published in Nature in the next weeks concluded
that Benveniste’s team did not reproduce the initial results and
that the alleged results were mainly the result of an observer’s
bias and unawareness of statistical laws.
4
In summary, the au-
thors of the report concluded that the disputed results were the
consequence of self-delusion. All the experiments performed for
years by Benveniste’s team were not taken into account or were
dismissed as statistically flawed. During the weeks that followed
the investigation, the debate raged in the correspondence pages
of Nature and other leading journals. Although most scientists
remained skeptical about the claims of Benveniste, the role of
Nature and its Editor was criticized. This debate was popularized
in mainstream press as the “memory of water” affair.
In the next years after publication, attempts to reproduce the
results of the Nature’s article produced negative,
5
ambiguous,
6
and positive
7-9
reports. This scientific saga is now considered as
a classic example of scientific controversy and is often described
as a conflict between “pseudo sciences” and mainstream science.
The different opinions have been polarized in two irreconcilable
camps, and no certainty has emerged. The only conclusion that
can be held for certain is that Benveniste’s work illustrates the
difficulty in exploring the fringes of science. Although the article
of 1988 has never been retracted, the role played by Nature has
probably hampered the efforts of some scientists who wished to
explore this controversial issue.
Today, for many scientists, the affair of the “memory of water”
is only an example of poor-quality science, and most of them
think that the Nature’s report marked the end of Benveniste’s
research on high dilutions. Actually, after the episode of 1988,
Benveniste abandoned the criticized basophil model for other
biological models that he hoped to be more easily convincing.
Several biological models were thus explored during the next
years and the most impressive results were obtained successively
with the isolated heart model and the in vitro coagulation
model. However, after all these years, the failure to convince
other scientists of the existence of a new research area is now
obvious. Even if conservative forces do exist in science as in
other human activities, Benveniste himself admitted that the
reproducibility of the experiments was a real concern. The thesis
developed here argues that the effects observed by Benveniste
and coworkers and supposed to be related to “memory of water”
were in fact related to a phenomenon that occurred unbeknown
to them. We will see that the obstacles encountered by Ben-
veniste’s team were not trivial and paradoxically could be a key
to understanding these puzzling results.
In this text, we first focus on the results obtained with the
isolated heart model. These results were obtained by Ben-
veniste’s team from 1990 to 1999. Less famous than the results
on basophils, these findings were nevertheless published as ab-
stracts and posters at international congresses. Moreover, Ben-
veniste’s team set up many “public demonstrations” aimed to
convince other scientists of the reality of the “memory of water.”
Most of these demonstrations were carefully designed with a
written protocol, and after completion, the participants received
a detailed report with the experimental raw data. Therefore,
considerable data were available that could be reanalyzed. These
experiments—and the whole story of the “memory of water”—
have been described in detail.
3
In a second step, a formal descrip-
tion of these results based upon the principles of the relational
interpretation of quantum physics will be proposed.
BRIEF DESCRIPTION OF THE BIOLOGICAL MODEL
(THE LANGENDORFF APPARATUS)
The Langendorff apparatus allows maintaining a live rat or
guinea pig heart while different parameters (beat frequency, cor-
onary flow, muscular tension) are recorded before and after in-
jection of pharmacological agents in the perfusion circuitry. As
reported in the section The successive devices used to “inform water”,
coronary flow was the most noticeable parameter because it
responded to “informed water” with large changes. In the Lan-
gendorff system, the perfusion liquid is forced to pass through
the coronary arteries before flowing outside the heart. Therefore,
any change in the diameter of the coronary arteries modifies the
flow rate. In the apparatus set up by Benveniste and coworkers,
the liquid that flowed out from the heart was recovered into a
series of tubes (one tube per minute), and the volume of liquid
was measured in each tube. Each run (test of one sample) was
performed in approximately 30 minutes. For each test, the pro-
file of the changes in coronary flow with time could be drawn
and the maximal coronary flow change was calculated as a per-
centage of basal flow. An empirical limit of 10% of the maximal
coronary flow change allowed the separation of “active” samples
(⬎10%) from “inactive” samples and background noise (both
⬍10%).
THE SUCCESSIVE DEVICES USED TO “INFORM”
WATER
In the early experiments with the Langendorff system, J. Ben-
veniste and coworkers observed some changes in the cardiac
physiological parameters after injection of highly diluted phar-
macological agents (eg, histamine) into the perfusion liquid.
10
In
fact, the most impressive effects were obtained in the coronary
flow. The effects observed in the other parameters, such as car-
diac frequency, were more erratic. Therefore, Benveniste’s team
focused its attention on the changes in coronary flow, and this
parameter became the main criterion for detecting the effects of
“informed” water. The major advantage of the Langendorff
186 EXPLORE May/June 2012, Vol. 8, No. 3 Explaining Signals in “Memory of Water” Experiments
model over the basophil model was the possibility to demon-
strate the biological effect to a public in real time. Indeed, the
20% to 30% change from the baseline flow was directly visible in
the series of tubes that collected the physiological solution from
coronary circulation. However, the criticism of possible contam-
ination of samples containing high dilutions by active molecules
remained.
In 1992, Benveniste claimed that he was able to “transfer” to
naive water, through a low-frequency amplifier, the biological
activity of a biologically active solution placed in an electric coil.
This “electromagnetic transfer” could be applied to naive water
contained in a sealed vial. Therefore, explaining the observed
effects by contamination was less relevant.
A further step was achieved in 1996 when Benveniste used a
personal computer with a sound card to “record” and to store as
a digital file the “activity” of the biological sample placed in the
electric coil. The “replay” to a naive water sample, which was put
inside an electric coil wired at the output of the sound card,
could then be performed. Positive results similar to those ob-
served with high dilutions were obtained. If reproducible, these
experiments offered huge possibilities in the area of diagnosis,
therapeutics, and fundamental biology. Benveniste then coined
the expression “digital biology.” The next successful step was
accomplished by placing the electric coil (which diffused the
electromagnetic “information”) directly around the column of
physiological liquid that perfused the isolated heart. Therefore,
no intermediary water sample was necessary, and the experiment
could be piloted directly from the computer. The electromag-
netic field of the electric coil was the unique link between the
electronic device and the biological system. The contamination
argument, which had been frequently proposed to explain these
results, was thus definitively discarded.
However, despite these successive improvements, a problem
persistently plagued the demonstrations aimed to provide
“proof” of the reality of the “memory of water.” This odd phe-
nomenon is described in the context of a “public demonstra-
tion” in the next section.
AN EXAMPLE OF “PUBLIC DEMONSTRATION”
As a representative “public demonstration,” an experiment per-
formed in September 1997, is described in detail; other similar
examples have been described elsewhere.
3
For this experiment
designed as a “proof of concept” to demonstrate the validity of
“digital biology,” Benveniste and coworkers performed the re-
cording step in a foreign laboratory and tested the coded com-
puter files in their laboratory. Two types of samples were “re-
corded”: water alone for negative controls and a solution of
calcium ionophore for positive controls. Calcium ionophore is a
compound that activates cells by causing calcium to permeate
the cell membrane.
In the foreign laboratory, J. Benveniste first recorded two
computer files (in .WAV sound file format) named “File water
initial” and “File iono initial” from tubes containing water or a
solution of calcium ionophore placed successively in the electric
coil. Then, the members of the foreign laboratory themselves
recorded three files named “File water 1” through “File water 3,”
followed by three files named “File iono 1” through “File iono
3.” Finally, Benveniste recorded two files named “File water
final” and “File iono final.”
The four files recorded by Benveniste were kept “open”; the
six files recorded by the guest laboratory members were copied
randomly into 10 files (some of the initial six files were thus
copied into more than one file) and then coded blind from 1 to
10. Then, Benveniste’s team moved back to their laboratory for
the second step. The recorded files were then “played” to naive
water and the resulting “imprinted” water was perfused to iso-
lated hearts. It is important to note that each file to be tested was
coded again (including the four “open-label” files) by a member
of Benveniste’s laboratory and then given to the experimenter
assigned to the tests. When all the experiments had been per-
formed, the results were sent to the scientists of the foreign
laboratory, who compared the effects observed on the apparatus
with the initial records that had been given a code-name. The
results are shown in Table 1.
We see in Table 1 that the differences between the samples
that appeared to be “active” or “inactive” were clear-cut. How-
ever, after code breaking, the “guessing game,” although based
on coherent experimental results, was not better than random.
Furthermore, some computer files, which had been simply du-
plicated, gave contradictory results (see, for example, “File iono
#1”). Note also that six files were found to be “active” in this
experiment, whereas only five “active” files were included in the
set of 10 files. In other public experiments where the number of
“active” samples had been fixed by the protocol and was public,
the correct number was obtained.
3
In other words, the experi-
mental results revealed only information that was available and
not hidden by the coding. It is important to insist again that in
this experiment described in Table 1, all runs were performed
blind for the experimenter.
Such disappointing results were the rule for public demonstra-
tions.
3
Faced with these results, J. Benveniste did not conclude
that these experiments “falsified” (in the Popperian sense) the
“memory of water” hypothesis, but he considered that the ex-
perimental conditions needed improvements to produce more
convincing results. He successively suggested several possible
causes that could disturb the results, such as electromagnetic
pollution, quality of water, “remnant” memory in the apparatus,
and the jump of “activity” between tubes. How explain, how-
ever, why the results with open-label samples, which were man-
aged in parallel with the blinded samples, were as “expected” and
were not affected by the supposed ad-hoc disturbances? Never-
theless, the determination of Benveniste to continue these ex-
periments was understandable: indeed, the experimental system
provided coherent outcomes, passed blind in-house tests, and
produced results that supported the initial hypothesis.
DISTRIBUTION OF THE SIZES OF THE BIOLOGICAL
EFFECTS ATTRIBUTED TO “INFORMED WATER”
Obviously, the causal relationship between biological effects
(changes in coronary flow greater or ⬍10%) and their supposed
causes (samples of “specifically informed water”) was disturbed
in some circumstances. As a first approach in the reanalysis of
these results, the abstracts of congress communications on the
Langendorff model written by Benveniste’s team were collec-
187
Explaining Signals in “Memory of Water” Experiments EXPLORE May/June 2012, Vol. 8, No. 3
ted.
10-20
Then, all results were extracted, and they were classi-
fied regardless of the methods supposed to have “informed”
the water samples and regardless of the biological molecules
supposed to have left an “imprint” in water. It appeared that
the distribution of the biological effects for samples reported
to be active was unimodal (modal class with 20%-30% of the
maximal change in coronary flow). This distribution is de-
scribed in Figure 1.
The unimodal distribution was confirmed by examining other
experiments performed by Benveniste’s team but not commu-
nicated at congresses. Particularly, large-scale demonstrations
confirmed this notion of a unimodal distribution for alleged
“active” samples (data not shown). For example, homeopathic
granules (“acetylcholine” or controls) were dissolved in water
and the solution applied to the Langendorff system after proper
dilution in physiological liquid. Again, a maximal change in
coronary flow of approximately 20% to 30% was measured for
“active” samples.
Such uniformity in results was rather astonishing. Indeed, the
processes used to “inform” water were very different. What is
common among high dilutions, direct “electromagnetic trans-
fer” from a biological sample, “electromagnetic transfer” from a
stored file, and transfer of the “biological activity” of homeo-
pathic granules to water? Moreover, a variety of electronic de-
vices have been used, particularly electric coils with various tech-
nical characteristics.
Thus, the dynamic range of the “measure apparatus” used to
evidence “informed water” appeared to be extremely large for
the “input” but was nevertheless associated with a monotonous
response for the “output.” In other words, the effect size ap-
peared to be binary (“it works” or “it does not work”) and not
continuous. What seemed to be important for the outcome was
the fact that it was the sample that was labeled as “inactive” or
“active” and not the specific physical process used to supposedly
“inform” the water.
CORRELATIONS OF PARALLEL MEASURES ON TWO
LANGENDORFF DEVICES
Without making any assumption on causal relationships, we
pursued this analysis by studying which events were observed
together. For this purpose, a set of experiments performed with
the Langendorff model between 1992 and 1996 was analyzed.
Indeed, two parallel Langendorff apparatus (here named A and
B) were used during this period in Benveniste’s laboratory. This
double apparatus was used to confirm the results of measure-
ments, particularly for the public demonstrations (note that
such double apparatus is seldom used in “normal” research).
These duplicate measures gave us the opportunity to analyze
the correlations between the measures obtained with apparatus
A and B. All the measures performed in duplicate during large-
scale experiments, essentially public demonstrations, were col-
Table 1. Example of a “Public Demonstration” with Participating Outside Observer Performed by Benveniste and Coworkers in September 1997
Tested Files
a
Size of Biological Effect (%)
(in Increasing Order) Mean ⫾SD
Unblinding of Blinded
Files
Expected Biological
Effect
Blind files with participating outside
observer
File 4 4.3 ⫾0.2 (n ⫽4) File “water,” 1 Yes
File 2 4.7 ⫾1.6 (n ⫽7) File “water,” 3 Yes
File 7 5.0 ⫾2.4 (n ⫽5) File “iono,” 1 No
File 8 5.1 ⫾4.0 (n ⫽4) File “iono,” 1 No
File 1 16.2 ⫾9.1 (n ⫽11) File “water,” 2 No
File 3 17.1 ⫾10.8 (n ⫽5) File “iono,” 3 Yes
File 10 20.3 ⫾15.8 (n ⫽6) File “iono,” 2 Yes
File 5 21.3 ⫾11.3 (n ⫽4) File “water,” 1 No
File 6 22.9 ⫾10.3 (n ⫽4) File “water,” 3 No
File 9 26.9 ⫾16.2 (n ⫽6) File “iono,” 1 Yes
In-house blind files
File “water” (initial) 3.1 ⫾0.3 (n ⫽5) — Yes
File “water” (final) 2.3 ⫾1.2 (n ⫽6) — Yes
File “iono” (initial) 24.0 ⫾4.5 (n ⫽5) — Yes
File “iono” (final) 25.2 ⫾15.0 (n ⫽7) — Yes
Positive control
Iono 1
mol/L 36.7 ⫾18.5 (n ⫽8) — Yes
a
The participating outside observer produced 10 blind files and four open-label files; the four open-label files were nevertheless in-house blinded before
measurements (see text for details). After completion of the measurements, the results were sent to the participating outside observer, who compared the two series
(labels of files and size of biological effects). The biological effect observed was the maximal change in coronary flow (expressed in percentages) using a Langendorff
apparatus. Despite the coherence of the results for each file, blind files were associated randomly with “effect” (size of biological effect ⬎10%) and “no effect”
(size of biological effect ⬍10%). In contrast, expected results were constantly obtained with the in-house blinded files. “Iono 1
mol/L” was a positive control of
calcium ionophore at “classical” concentration and was the last sample tested on a heart preparation. The number within parentheses in the second column is the
number of experiments performed for a given sample (on different heart preparations).
188 EXPLORE May/June 2012, Vol. 8, No. 3 Explaining Signals in “Memory of Water” Experiments
lected (these experiments have been described in detail
3
). All the
results (574 pairs of measures) were pooled regardless of the
process used to “inform” water (ie, high dilutions, direct electro-
magnetic transfer, transfer of “biological activity” previously re-
corded in computer files) or according to the biological mole-
cules involved (ie, acetylcholine, ionophore, ovalbumin).
Figure 2A indicates that the measures of the biological param-
eter (ie, maximal changes in coronary flow) obtained with appa-
ratus A and B were correlated. After log transformation (Figure
2B), we observe a more regular dispersion of the experimental
points. Indeed, the variable is a flow, which has the mathemat-
ical form K
R
2
; after log transformation, the mathematical
expression is linearized to K=
R (K and K=are constants). The
change in R, the coronary artery radius, is the real random vari-
able, which is indirectly measured by flow measurements.
In Figure 3A and 3B, the distributions of the measure values
are represented when the first value was ⬍10% (Figure 3A) or
⬎10% (Figure 3B). We observe that when the value measured on
apparatus A is ⬎10%, the probability to obtain a value ⬎10% on
apparatus B is high (this is symmetrical for measures ⬍10%).
This is another way to evaluate the degree of correlation for the
results obtained with the two devices. These samples were not
blinded between the first measure with A and the second mea-
sure with B.
The distributions of the measures in Figure 3A and 3B con-
firm the existence of two “states” of the biological model, each
with its own modal value (“inactive” and “active” states). The
very large dynamic range of the “measurement device” and the
monotonous profile of “active” samples observed above are con-
firmed here with a modal value of 20% to 30%.
Mixing results from a variety of experimental settings (with
various devices for water “information” and “transmission” of
different molecules) could be criticized. However, one could
argue that despite the combination of results obtained in heter-
ogeneous experimental conditions, the robustness of the uni-
0
5
10
15
20
0-10% 10-20% 20-30% 30-40% 40-50%
Number of values
Biological changes associated to "active" samples
(% of maximal variation of coronary flow)
Figure 1. Distribution of the size of the biological effects in different
experimental conditions (Langendorff system). Eleven communications
to congresses written by Benveniste’s team were analyzed.
10-20
Mean
values of active samples (maximal change in coronary flow expressed
as percentages of basal value) were extracted, and their distribution
was studied. Inactive controls (not represented in this figure) were all
in the range of 0% to 10%. Although the processes used to “imprint”
water were very different, the Langendorff system appeared to be able
to detect them, thus exhibiting a surprisingly large dynamic range for
input and a narrow range for output.
A
0
10
20
30
40
50
0 1020304050
Measure with device A
Measure with device B
B
1
10
100
1 10 100
Measure with device A
Measure with device B
Figure 2. Correlations of measurements performed in parallel with
the two Langendorff apparatus. A systematic analysis of large-scale
experiments from 1992 to 1996 (including mainly “public demonstra-
tions”) produced 574 pairs of measures. These duplicate measures
are plotted in A(scales limited to 50% for clarity). The log transform
of the values from Aare plotted in B.
189
Explaining Signals in “Memory of Water” Experiments EXPLORE May/June 2012, Vol. 8, No. 3
modal distribution of the size of the effects of the “active” sam-
ples is in favor of a unique alternative explanation.
It is important to note that log transformation was not used by
Benveniste and coworkers. Therefore, the symmetrical aspect
(Gaussian) of the two populations of effects (“inactive” and “ac-
tive”), which is evidenced in Figure 3A and 3B only after log
transformation, is an indirect evidence of the relevance of the
reported results. It is also important to note that the aim of the
public demonstrations performed by Benveniste was not to
prove a correlation between apparatus A and B but to verify the
fit of the results with the initial samples supposed to be “inac-
tive” or “active.” Therefore, these correlated duplicate measures
did not draw any particular attention and have not been high-
lighted by Benveniste. Nevertheless, I think these results are
particularly important because they indicate that coherence was
present in these puzzling experiments despite apparent discor-
dances. The aim of the following analyses is to understand when
correlations persist and when they are broken.
IN WHICH EXPERIMENTAL CONDITIONS DID
CORRELATIONS VANISH?
Experimental Situations with Correlations
We have seen in the last section that the values of the mea-
sures with the two parallel devices A and B were correlated.
Thus, when the value on apparatus A was ⬎10%, the proba-
bility to obtain a value ⬎10% on apparatus B was high (this
was symmetrical for measures ⬍10%; Figure 3A and 3B; sit-
uation 1 in Table 2).
In Table 2, other experimental situations are described. Thus,
in situation 2, an in-house coding was performed between the
first and the second measurements (both measurements were
performed with the same apparatus). In these in-house blind
conditions, correlations were also obtained (in this case, we do
not worry about the initial “label” of the sample, ie, samples
supposed to be active or controls). Again we observe that the
results of the measurements (⬎10% or ⬍10%) were concordant
despite the blinding. In situation 3, open-label samples sup-
posed to be “inactive” and “active” were prepared. Measure-
ments were performed after in-house blinding and correlations
were also observed. Therefore, in these first three situations with
open-label experiments or in-house blind experiments, the sta-
tistical difference in effects between “inactive” and “active” sam-
ples is very significant.
Experimental Situation without Correlation
The crucial issue is depicted by the experiments of situation 4
with coding of the samples by a participating outside observer
(an experimental situation comparable to the public demonstra-
tion described above). When all measurements had been per-
formed by the experimenter on the Langendorff apparatus, the
results were sent by Benveniste’s team to the participating out-
side observer, who held the code of the samples and who com-
pared the two series (biological effects and labels of the corre-
sponding samples). In this situation, the biological effects
(⬍10% and ⬎10%) were distributed at random according to the
initial label (“inactive” or “active” samples; Table 2). In sum-
mary, correlations were evidenced either in open-label experi-
ments or in-house blind experiments; in sharp contrast, in blind
experiments involving a participating outside observer, the cor-
relations vanished.
WHAT HAS BEEN MODIFIED IN THE EXPERIMENTAL
PROTOCOL TO ADAPT IT TO BLIND PUBLIC
DEMONSTRATIONS?
All bench scientists know that tiny modifications in an experi-
mental protocol could have unexpected negative consequences
on outcomes. Therefore, we have to wonder what has been
0
5
10
15
20
25
30
35
40
001011
Measure device B
(for measures device A < 10%)
Percentages
A
93% 7%
0
5
10
15
20
25
30
35
110100
Measure device B
(for measures dev ice A > 10%)
Percentages
B
11% 89%
Figure 3. Concordance of outcomes between identical apparatus A
and B for paired measurements. The distributions of the measures on
the two parallel Langendorff apparatus when the first measure was
⬍10% (A)or⬎10% (B) are described; samples were not blinded
between A and B measurements (same experimental points as in
Figure 2). The percentages on the figures indicate the proportions of
results below or above the 10% limit (n ⫽372 and n ⫽202 paired
measurements for Aand B, respectively).
190 EXPLORE May/June 2012, Vol. 8, No. 3 Explaining Signals in “Memory of Water” Experiments
modified between in-house and public experiments. The public
demonstrations set up by Benveniste’s team were generally per-
formed in two steps. In a first step, negative and positive samples
were produced (high dilutions, samples of “informed water” or
digital files) and were blinded with a code by an observer not
belonging to Benveniste’s team (whom we call a “participating
outside observer”). Some negative and positive samples were
kept open. In a second step, Benveniste’s team tested all blind
and open-label samples. When all measurements were com-
pleted, the results were sent (by fax or e-mail) to the participating
outside observer who checked the experimental outcomes and
the labels of the corresponding initial samples.
Therefore, a noteworthy modification to the initial protocol
was the checking of the two series after completion of experi-
mental data. We can now describe two experimental situations
that gave quite different results. In the first case, there was no
blinding or in-house blinding, and the expected correlations
were observed. In the second case, a participating outside ob-
server checked the two series after completion of the measures,
and the experiment was a “failure” (no significant correlations
were observed). Once more, it is important to emphasize that
usually samples kept open-labeled by the participating outside
observer were nevertheless in-house blinded (see the section An
Example of “Public Demonstration”).
Because the outcomes of the trials seem to strongly depend on
the people who performed the experiments and checked the
results, these experiments appear to be experimenter-dependent.
A similar unusual conclusion was already reported for these
experiments by a multidisciplinary team that was mandated to
assess the “digital biology” of Benveniste, as explained in the
next section.
THE END OF THE “MEMORY OF WATER”
HYPOTHESIS?
Laboratories able and willing to reproduce the experiments with
the Langendorff system were rare when Benveniste proposed to
replicate them. Therefore, Benveniste developed a new biologi-
cal model based upon the coagulation of plasma. An automated
robot analyzer was then set up to perform the coagulation ex-
periments with a limited intervention by the experimenter. The
different steps of the experiments were automatically performed:
random choice of computer files, “irradiation” of water samples,
distribution of samples and biological reagents in wells, and
quantification of coagulation (by optical density reading). In its
last version, the coagulation model was purely biochemical, with
coagulation of fibrinogen by thrombin in the presence of inhib-
itory “digital heparin.” The previous results of “digital biology”
and high dilutions were confirmed using this model, either man-
ually or with the automated analyzer.
The story of “digital biology” reached its highest point in 2001
when a multidisciplinary team of scientists attempted to repli-
Table 2. Reappraisal of Benveniste’s experiments with the Langendorff Apparatus: Concordant and Discordant Outcomes in Different
Experimental Conditions
Number of
Experimental
Points
% of Experimental Points
with Size of the Biological
Effects ⬍10%
% of Experimental Points
with Size of the Biological
Effects ⬎10% P-Value
a
Open-label experiments
Situation 1: apparatus A vs. apparatus B
b
Value ⬍10% after measurement with
apparatus A
n⫽372 93% (apparatus B) 7% (apparatus B) ⬍1⫻10
⫺83
Value ⬎10% after measurement with
apparatus A
n⫽202 11% (apparatus B) 89% (apparatus B)
In-house blind experiments
Situation 2: first vs. second measurements
of the same sample
Value ⬍10% after first measurement n ⫽50 96% (second measurement) 4% (second measurement) ⬍1⫻10
⫺13
Value ⬎10% after first measurement n ⫽28 7% (second measurement) 93% (second measurement)
Situation 3: “Inactive” vs. “active”
samples
“Inactive” samples n ⫽68 88% 12% ⬍1⫻10
⫺13
“Active” samples n ⫽58 19% 81%
Blind experiments with participating outside
observer
Situation 4: “Inactive” vs. “active”
samples
“Inactive” samples n ⫽54 57% 43% 0.25
“Active” samples n ⫽54 44% 56%
Percentages of concordant outcomes that are statistically significant are indicated in bold type.
a
Chi-square test.
b
See also Figure 3 for situation 1.
191
Explaining Signals in “Memory of Water” Experiments EXPLORE May/June 2012, Vol. 8, No. 3
cate Benveniste’s experiments on coagulation experiments by
using the automatic robot analyzer at the request of the United
States Defense Advanced Research Projects Agency. As reported
in an article describing the results of these experiments, some
effects that supported the concept of “digital biology” were ob-
served.
21
However, the experts in the multidisciplinary team did
not conclude that the effects of “digital biology” were real be-
cause they noticed that these effects could not be observed in-
dependently of the experimenter in Benveniste’s team assigned
to this experimentation. Interestingly, the authors concluded
that unknown “experimenter factors” could explain these odd
results but that a theoretical framework was necessary before
trying to apprehend them; they added: “Without such a frame-
work, continued research on this approach to digital biology
would be at worst an endless pursuit without likely conclusion,
or at best premature.”
21
Indeed, if the presence of some people is necessary to observe
a biological effect, are we still talking about water properties?
22
Taken together, all these results suggest strongly that the “mem-
ory” in the “memory of water” experiments was not located in
water. However, a puzzling question remains: how did a signal
nevertheless emerge from background noise? To answer this
issue, we propose to describe these experiments by using con-
cepts from relational quantum physics.
RELATIONAL QUANTUM PHYSICS
Description of the Relational Interpretation
In quantum physics, the evolution of a system is described by
the Schrödinger equation; it is called the evolution of the wave
function. This evolution is determinist, but this determinism is
not for events, as in classical physics, but for the probabilities of
these events. As a consequence, we cannot predict the outcomes
of individual measurements but only their probabilities.
The “measurement problem” has its roots in the discrepancy
between the linear superposition of the different states (as de-
scribed by the Schrödinger’s equation) and the actual measure-
ments, which always find the physical system in a definite state
(and not in a superposed state). Moreover, there is no indication
in the quantum formalism of a boundary between the micro-
scopic and the macroscopic world: the observers and their mea-
surement apparatus are themselves described by a determinist
wave function and for them also no precise results can be pre-
dicted for measurements, only probabilities. The purpose of the
different interpretations of quantum physics is to establish a
correspondence between quantum and classical reality.
The relational interpretation of quantum physics arose from a
reflection on the measurement problem.
23,24
To explicitly ad-
dress the measurement problem, Rovelli
23
supposes an observer
(named O) who makes a measurement of a parameter of a quan-
tum system S, which has two possible outcomes, namely “2”
and “1.”
In quantum physics, all the knowledge on a physical object
can be summarized by a “state vector” written as |
典, and the
state of the system S is described by the following state vector:
|
典⫽
␣
|↓典⫹

|↑典. This means that, after a measurement, the
probability to observe the result “2”is
␣
2
and the probability to
observe the result “1”is

2
(with
␣
2
⫹

2
⫽1;
␣
and

are
complex numbers).
In relational quantum mechanics (RQM), the state vector is
the information that a given observer possesses on a quantum
object. Therefore, a quantum description of the reality should be
always in reference to an observer, and there is no metaobserver
of the reality. In addition, there is no distinction between micro-
scopic and macroscopic worlds in RQM; all physical events are
quantum events and there is no “wave function collapse” (ie, the
wave function remains in a superposed state and the wave func-
tion collapse appears only in the states relative to a given ob-
server).
After a measurement, O observes either “2”or“1,” with the
respective probabilities
␣
2
and

2
. If we suppose that the out-
come of the measurement by O is “1,” then the evolution of the
system between t
1
and t
2
is:
ⱍ
典⫽
␣
ⱍ↓典⫹

ⱍ↑典→ⱍ↑典
We now suppose an observer P, who describes the system
formed by S and O, and its evolution (Figure 4). We suppose
that P has complete information on the initial state of this sys-
tem but does not make an observation of it during its evolution.
If the initial state vector of O is |init典, then the evolution of the
system S–O between t
1
and t
2
is described by P as:
共
␣
ⱍ↓典⫹

ⱍ↑典
兲
丢ⱍinit典→
␣
ⱍ↓典丢ⱍO↓典⫹

ⱍ↑典丢ⱍO↑典
where |O↓典and |O↑典are the state vectors of O who has observed
“2”or“1”, respectively. Therefore, the wave function describ-
ing the system S–O evolves into a superposition of two states:
the observer has obtained the result “2” in one state and the
result is “1” in the other state.
We conclude that, at the time t
2
, O and P make different
accounts of the same events. For P, the system is in a superposed
P
O
(GNU Free Documentation License)
Figure 4. The observer observed in the relational interpretation of
quantum mechanics. The observer O measures system S and P
observes O. What information does P have on O? (see text).
192 EXPLORE May/June 2012, Vol. 8, No. 3 Explaining Signals in “Memory of Water” Experiments
state after measurement; for O, a unique value has been ob-
tained. Thus, we have at our disposal two correct descriptions,
but they differ according to the observer considered. If, at a
time t
3
after t
2
, P interacts physically with S–O, then the state
vector of S–O “collapses” and only one “branch” is observed
by P. In each “branch,” P observes that the state of S and the
state of O are correlated. The state of P himself becomes
correlated with one possible state of S–O. The states of S, O
and P are correlated and S, O and P are said to be entangled.
This is what Rovelli names the “main observation” on which
the rest of the relational interpretation relies: “In quantum
mechanics different observers may give different accounts of
the same sequence of events.”
EPR Experiment and Relational Interpretation
The relational interpretation provides an interesting solution
to the EPR “paradox” by dissolving it.
25
The EPR paradox
considers two quantum objects in the following superposed
state: |
典⫽
␣
|↓典1|↑典2⫹

|↑典1|↓典2. The measurements on the
two objects are correlated (the two quantum objects are en-
tangled).
According to the orthodox interpretation of Copenhagen,
as soon as a result has been obtained after the first measure-
ment (for example “1”), then the value of the second object
is immediately fixed (“2” in this case). Therefore, the price to
pay for this interpretation is to suppose superluminal transfer
of information (nevertheless, this superluminal transfer can-
not carry useful information because the pairs of measures
“2
1
1
2
”or“1
1
2
2
” are obtained at random). Moreover, we
have to admit a “nonlocal” description of the world: a mea-
surement at one place in the universe could have immediate
consequences at another place, even if there is an astronom-
ical distance between them.
According to the Everett’s interpretation (as popularized by
De Witt), after measurement, the two possible pairs of results
(“2
1
1
2
” and “1
1
2
2
”) are both observed (there is no wave
function collapse) but in two different universes. The locality is
preserved in this interpretation, but the price to pay is the pro-
liferation of universes.
In the context of relational quantum physics, Rovelli and
Smerlak
25
pointed out that, in a system of two entangled quan-
tum objects, an observer never makes both measurements simul-
taneously. Indeed, the observer makes a measurement of the first
object and one “branch” of the superposed state of the first
object is selected at random. Then, the observer uses classical
means (at a speed below light velocity) to measure the second
quantum object which remains superposed (for an observer P) even
after the first measurement. The “branch” of the superposed
state of the second object is selected by taking into account the
past of the observer, that is, the result he or she has obtained after
the first measurement. Therefore, there is no superluminal trans-
fer of information and the principle of locality is preserved for
each observer. The price to pay for this interpretation is a weak-
ening of realism. The interpretation of Rovelli could be of some
help in our present issue with the notion of experimenter-depen-
dent outcome and with the absence of a distinction between
microscopic and macroscopic worlds.
DESCRIPTION OF BENVENISTE’S EXPERIMENTS
USING THE RELATIONAL INTERPRETATION OF
ROVELLI
Definitions
The purpose of the experiments to be described was to correlate
sample labels with apparatus outcomes. We summarize the sam-
ple labels as “inactive” and “active” (abbreviated as IN and AC,
respectively); the outcomes of the apparatus are “background
noise” and “signal” (symbolized as “2” and “1,” respectively).
We describe a minimal experiment where the effect of a
unique “sample” is observed. It should be clear that, according
to the previous analyses, we consider all samples physically in-
distinguishable. The only difference is between “labels,” namely,
the “properties” that samples are supposed to possess.
The experimenter A (Alice) observes, in a first step, sample
labels and apparatus and, in a second step, she assesses the con-
cordance of the paired observations (label IN with outcome
“2”; label AC with outcome “1”). If the statistical analysis of
repeated experiments indicates that the pairs are significantly
correlated, the experiment is considered a “success.” If there are
too many discordant pairs (defined as label IN with outcome
“1” or label AC with outcome “2”), the experiment is a “fail-
ure.”
We consider the point of view of an observer P as defined
previously, who describes Alice observing sample labels and
apparatus outcomes. This observer P has a complete knowledge
of the initial conditions of the system (Alice) and describes its
evolution, but he does not interact with it.
Open-Label Experiments
The evolution of the state vector of Alice when she observes the
label is:
ⱍ
A init典→ⱍ
A典⫽
␣
ⱍAIN典⫹

ⱍAAC典with
␣
2⫹

2⫽1
The evolution of the state vector of Alice when she observes
the apparatus outcome is:
ⱍ
A init典→ⱍ
A典⫽aⱍA↓典⫹bⱍA↑典with a2⫹b2⫽1
For an observer P, the state vector of A after completion of the
experiment (ie, observation of both label and outcome) is:
ⱍ
A典⫽(ⱍAIN 典⫹ⱍAAC 典)丢(ⱍA↓典⫹ⱍA↑典) (1)
N.B. For clarity, the coefficients associated with the vectors are
not indicated in this equation and in the following equations.
For the moment we admit that the signal (“1”) is observed by
Alice with a probability that is not negligible (we will detail in the
section Emergence of a Signal from Background Noise how the signal
could have emerged from the background). We have seen that
open-label experiments were considered as successes (Table 2).
In this experimental situation, statistically significant correla-
tions between IN and “2,” on one hand, and AC and “1,” on
the other hand, are observed. Therefore, the state vector of A
after an open-label experiment can be described as follows:
ⱍ
A典⫽ⱍAIN典ⱍA↓典⫹ⱍAAC典ⱍA↑典(2)
In the description of the evolution of the system from Equation
1 to Equation 2, it is as if the discordant pairs |AIN典|A↑典and
193
Explaining Signals in “Memory of Water” Experiments EXPLORE May/June 2012, Vol. 8, No. 3
|AAC典|A↓典were “filtered.” If we suppose that all samples tested in
these experiments are physically undistinguishable (only their
labels are different: AC and IN) and that trivial causal relation-
ships have been discarded, there is no reason to observe different
outcomes with the different samples. Yet, significant correla-
tions have been observed.
Therefore, the question is what or who decides for concordant
paired observations? If there is nothing in the history of the
universe that could explain the “decision,” then the choice
should be considered as free. As a consequence, we make the
assumption that concordant paired observations result from an
act of free will. This free will is exerted by Alice on her own state
as formalized by the evolution of the state vector |
A典. Alice’s
free will is based upon her a priori knowledge about the results of
experiments. Some comments on free will in the context of
quantum physics are added in the section Free Will and Quantum
Physics.
In-House Blind Experiments
In Equation 2, the order of the two observations (labels and
outcomes) has no consequence, eg, |AAC典|A↑典is equivalent to
|A↑典|AAC典. This formalism simply means that AC and “1,” on
one hand, and IN and “2,” on the other hand, are observed
together by Alice, regardless of the order of the observations.
This indicates that the two experimental situations: (1) Alice
observes the label before the outcome (open-label experiment)or
(2) Alice observes the outcome before the label (in-house blind
experiment), are formally identical. This description fits precisely
to Benveniste’s experiments where both situations led to signif-
icant correlations (Table 2).
Emergence of a Signal from Background Noise
Using the same formalism, we describe now the emergence of a
signal from background noise. Indeed, if Alice does not exert her
free will, there is no “filter,” and her state vector is equal to the
Equation 1 after development:
ⱍ
A典⫽ⱍAIN典ⱍA↓典⫹ⱍAAC典ⱍA↓典
High probability ⫹ⱍAIN典ⱍA↑典⫹ⱍAAC典ⱍA↑典
Low probability
In the absence of a “filter,” we have seen that the probability
of Alice observing a signal is low; background noise is most
probably observed. We have to remember that the numbers of
“inactive” and “active” labels have been defined by the experi-
mental protocol. If we suppose that there are as many “inactive”
as “active” labels, then the most probable states of Alice after
completion of the experiment are |AIN典|A↓典with a 50% proba-
bility and |AAC典|A↓典with a 50% probability (we consider that the
probability of the states |AIN典|A↑典and |AAC典|A↑典are negligible
but not equal to zero).
If Alice exerts her free will, “discordant pairs” are filtered and
the distribution of the probabilities between the different states
of Alice is changed because the state vector of Alice has evolved
to: |
A典⫽|AIN典|A↓典⫹|AAC典|A↑典(Equation 2). We see easily that
the probability associated with |AIN典|A↓典is again 50% (half of
labels are IN); the probability associated with |AAC典|A↑典, which
was negligible before filtration, is now increased to 50% (half of
labels are AC). In other words, a signal has been forced to emerge
from the background as a result of branch selection.
The result of this process is like a magnifying glass, which
enlarges some parts of the reality perceived by the observer.
Nevertheless, this “reality” can be shared with other observers. If
an observer E (Eve) interacts physically with Alice after comple-
tion of the experiment, the state vector that describes Alice and
Eve is:
ⱍ
AE典⫽ⱍAIN典ⱍA↓典ⱍEIN 典ⱍE↓典⫹ⱍAAC典ⱍA↑典ⱍEAC典ⱍE↑典
In both branches of the state vector, Eve agrees with Alice—after
several experiments—that there is (1) an emergence of a signal
from the background and (2) a statistically significant correlation
between labels and outcomes: samples supposed to be active are
associated with a signal, and samples supposed to be inactive are
associated with background noise.
It is important to note that we are not presently supporting
remote effects by free will. Indeed, remote effects suppose a
classical causal relationship involving forces or fields. In the
present case, the emergence of a signal is the consequence of
entangled states. In this sense, the relationship between labels
and outcomes could be considered as acausal.
In the context of Benveniste’s experiments, we propose that
the different and successive experimenters acquired skill by ma-
nipulating the biological systems and measurement devices (for
example by using “classical” stimuli). We can see this ability to
filter random outcomes as an extension of associative learning
from the classical world to the quantum world. This could ex-
plain why some experimenters (such as Alice) were more com-
petent to obtain the “expected” correlations.
Experiments with Participating Outside Observer
Let us consider now the case with a participating outside ob-
server B (Bob). Bob is introduced into the experimental design to
blind and control the results of Alice; he does not interact with
Alice during her measurements. The experimental process is
therefore modified. Thus, after completion of the measure-
ments, Alice sends the results she obtained (ie, which samples are
associated with signal) to Bob, who compares them with the
labels he blinded. As a consequence, the assessment of correla-
tions by Bob is a chief difference for this modified experimental
design. We have seen that this experimental design with a par-
ticipating outside observer was not successful because no signif-
icant correlations between labels and outcomes were observed.
The state vector of Alice and Bob in this experimental situation
can be described as:
ⱍ
AB典⫽
共
ⱍBIN典⫹ⱍBAC典
兲
丢
共
ⱍA↓典⫹ⱍA↑典
兲
(3)
⫽ⱍBIN典ⱍA↓典⫹ⱍBIN典ⱍA↑典⫹ⱍBAC 典ⱍA↓典⫹ⱍBAC典ⱍA↑典(4)
In this experimental situation, a signal emerges nevertheless
from background, but we have to wonder why no significant
correlations are observed. Indeed, it could be argued that Bob is
nothing else than a quantum object and that there should be no
obstacle for Alice’s free will to filter the state vector of Equation
4 to get |
AB典⫽|BIN典|A↓典⫹|BAC典|A↑典.
However, this reasoning cannot be held. The main argument
relies on intersubjective agreement, which is guaranteed by
194 EXPLORE May/June 2012, Vol. 8, No. 3 Explaining Signals in “Memory of Water” Experiments
quantum physics. Indeed, after completion of the measure-
ments, Bob assesses the correlations between labels and biolog-
ical outcomes independently of Alice. Bob observes a significant
proportion of discordant pairs (Bob is supposed to be unable to
filter them). Moreover, Bob cannot feel himself in a superposed
state. Therefore, when Alice and Bob come together, these dis-
cordant pairs must be present in the state vector that describes
the two observers. Indeed, it is impossible that significant corre-
lations exist (according to Alice) and do not exist (according to
Bob) simultaneously. Therefore, the only way to reach an inter-
subjective agreement is to suppose that the state vector
|
AB典⫽|BIN典|A↓典⫹|BAC典|A↑典is not possible and that only
Equation 4 (which includes all possible pairs, both concordant
and discordant ones) is the correct description of the intersub-
jective reality of Alice and Bob. In this shared reality, Bob and
Alice agree that no significant correlations were observed in the
series of experiments they performed.
FREE WILL AND QUANTUM PHYSICS
When two electrons interact (ie, they are next to each other),
their spins (up or down) align in opposite directions like small
magnets. Before the interaction, the state vector that describes
the state of the system composed of the two electrons is:
ⱍ
e1e2典⫽ⱍup典1ⱍdown典2⫹ⱍup典1ⱍup典2⫹ⱍdown典1ⱍdown典2
⫹ⱍdown典1ⱍup典2
After the interaction, the state vector is |
e1e2典⫽|up典1|down典2⫹
|down典1|up典2. We see that some branches of the superposed state
vector (branches with spins of the two electrons pointing in the
same direction) have been “filtered.”
In physics, there are conservation laws, and one of them re-
quires that, after an interaction of two electrons, the sum of their
spins must be zero (ie, they must have opposite spins). Whatever
happens to these electrons—they can be separated at great dis-
tances—their total combined spin must remain zero. If one mea-
sures “2” for the first electron of the entangled pair, then we can
hold for sure that the measure of the second electron will be
“1”: the electrons are said to be entangled. Therefore, we can
suggest that the notion of free will that we have introduced
previously plays a role analogous to a conservation law (at least
for a limited duration) by filtering some states and therefore
entangling quantum objects.
The use of the notion of free will in a quantum description
should not surprise. When an outcome is not determined by the
past history of the universe, it can be considered as a conse-
quence of free will. On this basis, Conway and Kochen
26
have
demonstrated that if we have (some) free will, then quantum
particles also have (some) free will. Free will, defined in this
sense, together with the central role of the observer, is an essen-
tial element of the description of quantum events by quantum
theory. Zeilinger insists on two freedoms in quantum experi-
ments: first, the freedom of the experimenter when he chooses
the setting of the measuring apparatus, and second, the freedom
of the nature, which gives an answer; detection with a measuring
apparatus can thus be considered as an “elementary act of cre-
ation.”
27
These views will certainly be important to consider for
a theory of free will and consciousness that remains to be writ-
ten.
ANSWER TO THE MAIN OBJECTION
The usual criticism to our description of Benveniste’s experi-
ments using quantum formalism can be summarized by this
comment: “How quantum superposition could be guaranteed at
ambient temperature? Indeed, the environment of the experi-
ment should lead to quick decoherence.”
This comment would be quite pertinent within the framework
of the Copenhagen interpretation of quantum physics, but in
the context of RQM, this criticism is irrelevant. Therefore, we
have to insist briefly on some particular issues in RQM.
In RQM, the only way to say that an event has occurred or has
not occurred is to index it. If I measure that the spin of an
electron is up, I cannot conclude that the spin of the electron is
up. I can only say that the spin of this electron as measured by me
is up. In other words, the actual outcome is real only with respect
to the observer. Therefore, in contrast with classical physics,
where there is a unique account of the reality by multiple ob-
servers, in RQM there are as many accounts as observers. The
comparison of results between different observers is also a phys-
ical process and coherence is guaranteed by RQM; as suggested
by Vecchi,
28
“reality” is the locus of intersubjective agreement.
As a consequence of the multiple possible observers, there is no
wave function collapse; more precisely, the wave function collapse
appears only in the states relative to a given observer and decoher-
ence in this context is the theory that explains how the actual-
ization of the variable occurred for this observer. RQM allows
describing not the physical world itself but rather the general form
of information that one system can obtain about another. This
description is expressed in the form of correlations (between
entangled objects and observers); describing correlations in the
physical world is precisely the exact definition of science.
This simple (and easily acceptable) restriction to our discourse
on the physical world by indexing an event to an observer has a
huge advantage: it allows overcoming some of the difficulties of
other interpretations of quantum physics (Copenhagen, Everett,
etc). As an example, the relativization of actuality to each ob-
server allows reconciling quantum physics with locality.
25
As
already said, one consequence of RQM is a weakening of real-
ism.
It is important to insist that RQM should not be confused
with the “relative state” interpretation of Everett where there is a
multiplicity of realities; in RQM, a quantum event refers only to
a pair of systems. These few comments can be completed by
further readings on RQM, which is outside the scope of the
present paper.
23-25
CONCLUSIONS
In summary, our reappraisal of Benveniste’s experiments has led
to the conclusion that the results were experimenter-dependent.
A model based on the relational interpretation of quantum
physics describes the characteristics of these experiments: (1)
experimenter-dependent emergence of a signal from back-
ground noise, (2) loss of correlations between the supposed
195
Explaining Signals in “Memory of Water” Experiments EXPLORE May/June 2012, Vol. 8, No. 3
“causes” and the observed effects in some defined experimental
conditions, and (3) difficulties for other independent teams to
produce a signal and significant correlations. Therefore, al-
though our hypothesis does not dismiss definitely the possibility
of “memory of water,” the experimenter-dependent entangle-
ment could be an attractive alternative interpretation of Ben-
veniste’s experiments.
Acknowledgments
The author would like to gratefully acknowledge the construc-
tive comments provided by the journal editor and the three
anonymous reviewers.
REFERENCES
1. Maddox J. Waves caused by extreme dilution. Nature. 1988;325:
760-763.
2. Davenas E, Beauvais F, Amara J, et al. Human basophil degranula-
tion triggered by very dilute antiserum against IgE. Nature. 1988;
333:816-818.
3. Beauvais F. L’Âme des Molécules—Une histoire de la “mémoire de l’eau.
Collection Mille Mondes (ISBN: 978-1-4116-6875-1); 2007. Available
at: http://www.mille-mondes.fr. Accessed March 4, 2012.
4. Maddox J, Randi J, Stewart WW. “High-dilution” experiments a
delusion. Nature. 1988;334:287-291.
5. Ovelgönne JH, Bol AW, Hop WC, et al. Mechanical agitation of
very dilute antiserum against IgE has no effect on basophil staining
properties. Experientia. 1992;48:504-508.
6. Hirst SJ, Hayes NA, Burridge J, et al. Human basophil degranulation
is not triggered by very dilute antiserum against human IgE. Nature.
1993;366:525-527.
7. Benveniste J, Davenas E, Ducot B. L’agitation de solutions haute-
ment diluées n’induit pas d’activité biologique spécifique. Comptes
rendus de l’Académie des sciences (Série 2). 1991;312:461-466.
8. Belon P, Cumps J, Ennis M, et al. Inhibition of human basophil
degranulation by successive histamine dilutions: results of a Euro-
pean multi-centre trial. Inflamm Res. 1999;48;suppl 1:S17-S18.
9. Brown V, Ennis M. Flow-cytometric analysis of basophil activation:
inhibition by histamine at conventional and homeopathic concen-
trations. Inflamm Res. 2001;50;suppl 2:S47-S48.
10. Hadji L, Arnoux B, Benveniste J. Effect of dilute histamine on
coronary flow of guinea-pig isolated heart. Inhibition by a magnetic
field. FASEB J. 1991;5:A1583.
11. Benveniste J, Aïssa J, Jurgens P, et al. Effects on the isolated heart of
water preexposed to a permanent magnetic field. FASEB J. 1992;6:
A425.
12. Benveniste J, Arnoux B, Hadji L. Highly dilute antigen increases
coronary flow of isolated heart from immunized Guinea-pigs.
FASEB J. 1992;6:A1610.
13. Benveniste J, Aïssa J, Litime MH, et al. Transfer of the molecular
signal by electronic amplification. FASEB J. 1994;8:A398.
14. Benveniste J, Jurgens P, Aïssa J. Digital recording/transmission of
the cholinergic signal. FASEB J. 1996;10:A1479.
15. Benveniste J, Jurgens P, Hsueh W, et al. Transatlantic transfer of
digitized antigen signal by telephone link. J Allergy Clin Immunol.
1997;99:S175.
16. Benveniste J, Aïssa J, Guillonnet D. Digital biology: specificity of
the digitized molecular signal. FASEB J. 1998;12:A412.
17. Benveniste J, Aïssa J, Guillonnet D. The molecular signal is not
functional in the absence of “informed” water. FASEB J. 1999;13:
A163.
18. Litime MH, Aïssa J, Benveniste J. Antigen signaling at high dilution.
FASEB J. 1993;7:A602.
19. Aïssa J, Litime MH, Attias E, et al. Transfer of molecular signals via
electronic circuitry. FASEB J. 1993;7:A602.
20. Aïssa J, Jurgens P, Litime MH, et al. Electronic transmission of the
cholinergic signal. FASEB J. 1995;9:A683.
21. Jonas WB, Ives JA, Rollwagen F, et al. Can specific biological signals
be digitized? FASEB J. 2006;20:23-28.
22. Beauvais F. Memory of water and blinding. Homeopathy. 2008;97:
41-42.
23. Rovelli C. Relational quantum mechanics. Int J Theor Phys. 1996;35:
1637-1678.
24. Laudisa F, Rovelli C. Relational quantum mechanics. In: Zalta EN,
ed. The Stanford Encyclopedia of Philosophy (Fall 2008 Edition). Avail-
able at: http://plato.stanford.edu/archives/fall2008/entries/qm-
relational/. Accessed March 4, 2012.
25. Rovelli C, Smerlak M. Relational EPR. Found Phys. 2007;37:427-
445.
26. Conway J, Kochen S. The free will theorem. Found Phys. 2006;36:
1441-1473.
27. Zeilinger A; 2006. Interview, Die Weltwoche, Ausgabe 48/05; English
translation. Available at: http://signandsight.com/features/614.html.
Accessed March 4, 2012.
28. Vecchi I. Are classical probabilities instances of quantum ampli-
tudes? Available at: arXiv:quantph/0206147v1. Accessed March 4,
2012.
196 EXPLORE May/June 2012, Vol. 8, No. 3 Explaining Signals in “Memory of Water” Experiments
