Loading...
Reload document 
Open in new tab Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 24 February 2020 doi:10.20944/preprints201810.0060.v2
“Memory of Water” Experiments Explained with No Role Assigned to Water:
Pattern Expectation after Classical Conditioning of the Experimenter
Francis Beauvais 1, *
1 Scientific and Medical Writing, 91 Grande Rue, 92310 Sèvres, France
* Correspondence: beauvais@netcourrier.com; Tel.: +33 6 68 36 58 36
ABSTRACT
The “memory of water” experiments suggested the existence of molecular-like effects without
molecules. Although no convincing evidence of modifications of water – specific of biologically-
active molecules – has been reported up to now, consistent changes of biological systems were
nevertheless recorded. We propose an alternate explanation based on classical conditioning of the
experimenter.
Using a probabilistic model, we describe not only the biological system, but also the experimenter
engaged in an elementary dose-response experiment. We assume that during conventional
experiments involving genuine biologically-active molecules, the experimenter is involuntarily
conditioned to expect a pattern, namely a relationship between descriptions (or “labels”) of
experimental conditions and corresponding biological system states.
The model predicts that the conditioned experimenter could continue to record the learned pattern
even in the absence of the initial cause, namely the biologically-active molecules. The phenomenon
is self-sustained because the observation of the expected pattern reinforces the initial conditioning.
A necessary requirement is the use of a system submitted to random fluctuations with
autocorrelated successive states (no forced return to the initial position). The relationship recorded
by the conditioned experimenter is, however, not causal in this model because blind experiments
with an “outside” supervisor lead to a loss of correlations (i.e., system states randomly associated
to “labels”).
In conclusion, this psychophysical model allows explaining the results of “memory of water”
experiments without referring to water or another local cause. It could be extended to other
scientific fields in biology, medicine and psychology when suspecting an experimenter effect.
Keywords: Experimenter effect; “Memory of water”; Classical conditioning.
1. Introduction
The controversy over the “memory of water” that burst in 1988 continues to maintain in the
shadow the whole story of Benveniste’s experiments that extended over 20 years from 1984 to
2004.1 Admittedly these claims were anything but insignificant: the experiments presented in Nature
suggested the existence of molecular-like effects in the absence of molecules.2 The authors of this
article stated that dilutions of biologically-active molecules beyond the limit defined by the
Avogadro number had nevertheless a biological effect.
© 2020 by the author(s). Distributed under a Creative Commons CC BY license.
1/20
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 24 February 2020 doi:10.20944/preprints201810.0060.v2
The violence of the controversy had most probably its roots in the “two centuries of observation
and rationalization” that these results were supposed to reconsider.3 Because the idea that a
“structuration” of water could mimic the effects of biologically-active molecules was considered
impossible, the inevitable conclusion was that the experiments were flawed. As a consequence,
there was no place for an alternate theoretical framework that would consider these results, without
involving water and its alleged “memory”. The fact that this study could support homeopathy, also
highly controversial, was another reason for this strong opposition. It is out of the scope of this
article to describe this controversy; details on the debate and disputed experiments can be found
elsewhere.1, 4, 5
According to the judgement of many scientists, there was nothing to explain in these experiments
as there was no scientific facts, only poor science. Therefore, the report of Nature’s investigation in
Benveniste’s laboratory has been generally considered to put the last word to the public debate.6, 7
Nevertheless, some authors reported modifications of physical parameters of highly-diluted
solutions or proposed different theoretical frameworks.8-13 How the specificity of the initial
molecule could be conveyed through the successive dilutions remained however unanswered in
these various theoretical frameworks. Moreover, correlations of changes of water parameters and
corresponding changes of a biological model have not been described up to now.
My purpose in this introduction is not to fuel this debate again but simply to structure the
arguments from both sides to explain why Benveniste failed to convince his peers. Indeed, after
the basophil model described in 1988 in Nature’s article, other experimental models, mainly isolated
rodent heart and plasma coagulation, were developed by Benveniste’s team. Experimental data
accumulated seemingly in favor of a role of water for storing information on molecules in
solution.14-20 During this period, Benveniste made a step further by stating that molecular
information could be “imprinted” in water through electromagnetic fields (1992) as in a magnetic
tape 15 and could be even digitized (1995).20 At this occasion, he coined the expression “digital
biology”.20, 21
In Table 1, arguments from Benveniste’s experiments in favor of or against “memory of water”
are summarized. The arguments in favor of “memory of water” are mainly the observation of
“activated” states of the biological systems associated to test samples “imprinted” with different
methods and the apparent specificity of the biological effects. The arguments against “memory of
water” are mostly difficulties to reproduce the results by other teams and the absence of a
compelling theoretical framework. There is also another reason – less known – that prevented
Benveniste to go further in his quest of the perfect experiment that would be totally convincing.
This reason was a stumbling block that was more particularly highlighted during public
demonstrations where colleagues from other teams were invited to supervise proof-of-concept
experiments. The role of these outside supervisors was to produce “inactive” and “active” test
samples (water samples with high dilutions or “imprinted” water; computer files for digital biology)
and to mask them under a code number. After the outside supervisors had left, the coded samples
were tested by Benveniste’s team. When all measurements were completed, the results were sent
to the supervisors who assessed the rate of success by comparing for each run the measured system
state and the corresponding “label” (unbeknown of the experimenter who did the test). These
proof-of-concept experiments systematically failed in the sense that “activated” states were always
randomly distributed between test samples with “inactive” and “active” labels.1, 22, 23
To explain these troublesome failures, Benveniste proposed many post hoc explanations (e.g., water
contaminations, interferences with external electromagnetic fields, “jumps of activity” from one
test sample to another, human errors for sample allocation).1 Despite further improvements in
devices and procedures to prevent these disturbances, the weirdness persisted. The important
point, however, is that these possible external disturbances did not account for “successful” results
with open-label test samples even in blind conditions with an “inside” supervisor or an automatic
device (more precise definitions of “inside” and “outside” supervisors are given later).
2/20
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 24 February 2020 doi:10.20944/preprints201810.0060.v2
In this article, I propose to simultaneously take into consideration Benveniste’s experiments and
to abandon the “memory of water” hypothesis. A theoretical framework is constructed where these
experiments are related to an experimenter effect, which is the consequence of a previous classical
conditioning of the experimenter. In this setting, all test samples are nothing more than controls
(or placebos) and the different procedures to “imprint” water samples are nothing more than
rituals. The proposed model describes all features of Benveniste’s experiments: emergence of an
“activated state” of a biological system without local cause, correlations between “labels” and
system states, and mismatches of outcomes in blind experiments with an outside supervisor. No
role is attributed to water or another local cause but now the attention shifts to the experimenter.
The proposed experimenter effect is original and could have consequences beyond the “memory
of water” controversy. Therefore, considering Benveniste’s experiments only as an example of
specious science misses the point and prevents from seeing what these experiments could teach us.
The price that the proponents of “memory of water” have to pay is abandoning the initial
hypothesis (i.e., a molecular-like effect without molecules). For the opponents, the price to pay is
to accept that these weird experiments – admittedly misinterpreted by their authors – had
nevertheless a scientific interest.
a
Table 1. The arguments for and against molecular-like effects without molecules in Benveniste’s
experiments.
Arguments for
Arguments against
• Emergence of an “activated” state of biological
models mimicking the effect of biologically-active
molecules a
• Not compatible with current scientific knowledge
on water (e.g., very short half-lives of chemical
bonds between water molecules)
• Emergence of a relationship between
experimental conditions and states of system
• Specificity of the molecular-like effects b
• Consistency of the results with different
• Not compatible with current scientific knowledge
on biochemical interactions (e.g., law of mass
action)
• No compelling theoretical framework
experimenters, biological systems and procedures
• Difficulties to reproduce the results by other
• Successful tests in blind experiments with
teams
local/inside supervisor or automated devices. c
• Proximity with homeopathy
• Loss of correlations in blind experiments with an
outside supervisor. c
In “memory of water” experiments, water samples are supposed to induce a biological activity although
the biologically-active molecules have been removed via extensive serial dilutions (“high dilutions”) or
after water samples have been “imprinted” through electromagnetic fields using different devices
(“electronic transmission” or “digital biology”).
b Water samples supposed to have been “imprinted” apparently retained the specificity exhibited by the
original molecules (“imprints” of biologically-inactive molecules were inactive even if their structure was
close to biologically-active molecules).
c See definitions of “inside” or “outside” supervisors in section “Consequences of blind experiments on
correlations”.
2. Classical conditioning during ordinary dose-response experiments
Classical conditioning (or Pavlovian conditioning) is a well-known associative learning process.24
We briefly describe classical conditioning with a typical example before making a parallel with an
3/20
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 24 February 2020 doi:10.20944/preprints201810.0060.v2
experimenter who handles an experimental system. Classical conditioning supposes first an
“unconditioned stimulus” that produces an “unconditioned response” in an organism. In the well-
known example of Pavlov’s dog, smelling or tasting food (unconditioned stimulus) induces
salivation (unconditioned response). The purpose of the learning is to pair a “neutral stimulus” to
the unconditioned stimulus. In this case, a bell (neutral stimulus) systematically rings just before
food (unconditioned stimulus) is presented to the dog. Thus, the dog learns to associate the ring
of the bell and the coming of food. During this learning process, the former neutral stimulus
becomes a “conditioned stimulus”. Indeed, salivation (conditioned response) is now induced when
the bell rings. To be complete, we must add that no food is expected (no salivation) by the dog
when the bell does not ring. Thus, a relationship is established between the conditioned stimulus
(ring vs. no ring) and the conditioned response (salivation vs. no salivation).
The purpose of most in vitro or physiology experiments is to study the effect of a biologically-active
compound on a biological system. A dose-response is performed, meaning that the effect of the
compound at different concentrations (0, x, 2x, 3x, etc.) is evaluated. For simplifying, we suppose
that only one “active” condition versus one “inactive” condition (or control) is assessed during the
experiments. We suppose also that the biological system has only two states: “resting” state (not
different from background noise) and “activated” state (different from background noise).
In such experiments, one usually forgets that the biologically-active compound has not only a direct
effect on the biological system but also an indirect effect on another “biological system”, namely
the experimenter. Even with automated systems, there is always an experimenter who prepares the
experiment, records the outcomes and interprets them. Therefore, it is easy to take a step further
and to consider that during the repetitions of experiments, the experimenter unintentionally learns
to combine the experimental conditions with the states of the biological system. Thus, the
“inactive” condition (control) is associated to the “resting” state and the “active” condition
(biologically-active molecules at pharmacological concentrations) is associated to the “activated”
state.
After this classical conditioning process, the cognitive structures of the experimenter are changed.
The “labels” of the experimental conditions are associated to the respective system states: “inactive”
label is associated to “resting” state and “active” label is associated to “activated” state.
In the model that we construct, we posit that all samples to be tested are identical and are all
biologically inactive in the sense that they do not induce a local causal biological effect. Even if the
test samples are subjectively named “inactive” or “active” by the experimenter, it would be
impossible to distinguish one test sample from another one on physical bases; only their
identification with “labels” – in other words their meaning for the experimenter – is different. “Labels”
are nothing more than a short description for the experimenter about the expected state of the
experimental system. In Figure 1, the “labels” have the nonspecific names L1 or L2 that do not
presuppose to which experimental condition (“inactive” or “active”) they are respectively
associated by the experimenter.
It is important to underline that a relationship (direct or reverse) has a higher degree of abstraction
than its components (“labels” and system states). A relationship is similar to a pattern (or a shape)
that is thought in its wholeness, not as the simple sum of its individual components. In other words,
after conditioning, the experimenter expects an “image” (a continuous entity), not “pixels” (discrete
stuffs).
As the primary purpose of this article is to propose an explanation of Benveniste’s experiments,
we must underscore that the members of Benveniste’s laboratory routinely performed “classical”
experiments with genuine molecules. Therefore, classical conditioning can be easily assumed in this
context.
4/20
Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 24 February 2020 doi:10.20944/preprints201810.0060.v2
Figure 1. Expectation of patterns by the experimenter after classical conditioning. The two
“labels” (L1 vs. L2) and the two possible system states (“resting” vs. “activated”) define four
couples of outcomes (A). The “labels” have the nonspecific names L1 or L2 which do not
presuppose to which of the two experimental conditions (“inactive” or “active”) they are
respectively associated by the experimenter. After classical conditioning (i.e., Pavlovian
conditioning) with “conventional” experiments involving biologically-active molecules at
pharmacological concentrations, the two possible relationships expected by the experimenter are
named “direct” or “reverse” relationships with probabilities p and q, respectively (B). A
relationship has a higher degree of abstraction than its components and is similar to a pattern that
is thought in its wholeness, not as the simple sum of its individual components.
3. Consequences of the classical conditioning of the experimenter for future outcomes
In this section, we explore the probabilistic consequences of the classical conditioning of an
experimenter named F who records the state of the experimental system S (Figure 2). These
probabilistic consequences are described in three sequential steps:
Step 1: F-S taken as a whole. The state of the system S (“resting” or “activated” state) at the end of
the experiment is obviously a property of S. However, as previously said, F has been conditioned
and expects a pattern (direct/reverse relationship). The future outcome to be recorded by F is
therefore the combination of an abstract construct (pattern) and a physical variable (state of S).
This abstract construct is composed of “labels” which are arbitrarily chosen and do not correspond
to physical properties of test samples. As a consequence, the future outcome to be recorded by F
is neither an individual property of F nor an individual property of S but is a property of F and S
taken as a whole. It is important to underline that this future outcome is not the simple juxtaposition
of a first subevent that is a property of F and a second subevent that is a property of S. Indeed, F
and S constitute a new “entity” denoted F-S that cannot be dissociated.
5/20