"Maybe I should have thrown the data away"
but being a scientist and believing in his data he could not Jacques Benveniste, 1935-2004
The ‘memory of water’ is a popular phrase that is mostly associated with homeopathy and Jacques Benveniste  following his and others’ allergy research work . These research teams showed that solutes subjected to sequential physical processing and dilution show biological effects different from those apparent using just the water employed for the dilutions. The subject has drawn a lot of controversy with many scientists simply rejecting it outright without studying the evidence. The subject area has recently been the subject of a number of papers in the journal Homeopathy (July, 2007)c and has been reviewed . Although there is much support for water showing properties that depend on its prior processing (that is, water having a memory effect), the experimental evidence indicates that such changes are due primarily to solute and surface changes occurring during this processing. The experimentally corroborated memory phenomena cannot be taken as supporting the basic tenets of homeopathy although they can explain some effects .
The main evidence against water having a memory is that of the very short (~ps) lifetime of hydrogen bonds between the water molecules . Clearly in the absence of other materials or surfaces (see later), the specific hydrogen bonding pattern surrounding a solute does not persist when the solute is removed any more than would a cluster around any specified water molecule, or else water would not know which of its myriad past solutes took preference. A recent NMR study shows no stable (>1 ms, >5 μM) water clusters are found in homeopathic preparations . It should, however, be noted that the lifetime of hydrogen bonds does not control the lifetime of clusters in the same way that a sea wave may cross an ocean, remaining as a wave and with dependence on its history, but with its molecular content continuously changing. Also, the equilibrium concentration of any clusters are governed by thermodynamics not kinetics.
As applied to homeopathy, the 'memory of water' concept should also be extended to the memory of aqueous ethanol preparations. Addition of ethanol to water adds an important further area of complexity. Ethanol forms solutions in water that are far from ideal and very slow to equilibrate . Although usually considered a single phase, such solutions may contain several distinct phases  and more generally consist of a complex mixture dominated by water-water and ethanol-ethanol clusters, where hydrogen bonding is longer-lived than in water alone . They also favor nanobubble (that is, nanocavity) formation . Thus, the peculiar behavior of aqueous solutions (as mostly discussed on this page) is accentuated by the presence of ethanol.
The process of silica dissolution has been much studied [1109, 1207] ever since it was proven by Lavoisier over 200 years ago and fits with this argument. This may explain why glass is preferred over polypropylene tubes. It should be noted that dissolved silica is capable of forming solid particles with complementary structures (that is, imprints) to dissolved solutes and macromolecules and such particles will 'remember' these complementary structures essentially forever.
Water does store and transmit information, concerning solutes, by means of its hydrogen-bonded network. Changes to this clustering network brought about by solutes may take some time to re-equilibrate. Agitation (succussion) may also have an effect on the hydrogen bonded network (shear encouraging destructuring) and the gaseous solutes (with critical effect on structuring  and possible important production of structuring nanobubbles (nanocavities) ), and such effects may well contribute to the altered heats of dilution with such materials . Such mechanically induced hydrogen bond breakage may also give rise to increased (but low) hydrogen peroxide formation [1066 see equations] and such effects have been reported to last for weeks . It may be relevant to note that the presence of hydrogen peroxide can take part in and catalyze further reactions with other reactive species such as molecular oxygen and dissolved ozone [1066, 1069] (not often recognized but present in nanomolar amounts) which may well vary with the number of succussion steps and their sequence, which may offer an explanation for the changes in efficacy of homeopathic preparations with the number of dilutions . Also of note are the known effects of low concentrations of reactive oxygen species on physiological processes such as the immune response; with the recent discovery of the importance of low levels of hydrogen peroxide being particularly relevant .
Dilution is never perfect, particularly at low concentrations where surface absorption may well be a major factor, so that dilution beyond the levels that can be analytically determined remains unproven. Remaining material may be responsible for perceived differences between preparations and activity. Of course the water used for dilution is not pure relative to the putative concentration of the 'active' ingredient; even the purest water should be considered grossly contaminated compared with the theoretical homeopathic dilution levels. This contamination may well have a major influence, and itself be influenced by the structuring in the water it encounters. Although it does, at first sight, seem unlikely that solutes in diluted 'homeopathic' water should be significantly different from a proper aqueous control, it has recently been cogently argued that the concentrations of impurities can change during the dilution process by reactions initiated by the original 'active' material , and this process has been mathematically modeled .
A further consideration about 'the memory of water' is that the popular understanding concerning how homeopathic preparations may work not only requires this memory but also requires that this memory be amplified during the dilution; this amplification, necessitated by the increase in efficacy with extensive dilution, being even harder to explain. Samal and Geckeler have published an interesting, if controversial, paper  concerning the effect of dilution on some molecules. They found that some molecules form larger clusters on dilution rather than the smaller clusters thermodynamically expected. Just the presence of one such large μm-sized particle in the 'diluted' solution could give rise to the noticed biological action (of course, some such preparations may be totally without action, being without such clustered particles).a However, it remains to explain this particular phenomenon, which appears to disobey the second law of thermodynamics. A possible explanation is that such biologically-active molecules can cooperatively form icosahedral expanded water networks (ES) to surround and screen them by the formation of face-linked icosahedra, similar to as expected in the minimal energy related poly-tetrahedral Dzugutov clusters . So long as such an icosahedral network structure requires the help of more than one neighboring such cluster to stabilize its formation then, in more concentrated solution, the molecules dissolve normally. However, as they are diluted (typically beyond about one clathrate-forming group per twelve icosahedral water clusters; 3,360 water molecules) no neighboring such clusters are available and the clusters coalesce to form larger clusters of biologically-active molecules within their own ES-related water network (so releasing some of the water). This tendency for particle formation is ultimately due to the hydrophobic effect and the tendency to form a small surface with the water. Overall the balance is expected to be rather fine between water cluster stabilization and particle cluster stabilization.
Water is not just H2O molecules. It contains a number of molecular species including ortho and para water molecules, water molecules with different isotopic compositions such as HDO and H218O, such water molecules as part of weakly-bound but partially-covalently linked molecular clusters containing one, two, three or four hydrogen bonds, and hydrogen ion and hydroxide ion species. Apart from such molecules there are always adventitious and self-created solutes in liquid water. Distilled and deionized water contain significant and varying quantities of contaminating ions. Often the criteria for ‘purity’ is the conductivity, but this will not show ionic contaminants at nanomolar, or even somewhat higher, concentrations due to the relatively high conductivity of the H+ and OH- ions naturally present. Other materials present will include previously dissolved solutes, dissolved gasses dependent on the laboratory atmosphere, gaseous nanobubbles [500d], material dissolved or detached from the containing vessels , solid particles and aerosols (also dependent on the laboratory history) entering from the gas phase, and redox materials produced from water molecules  and other solutes produced on standing [509c] and homeopathic processing . Liquid water is clearly a very complex system even before the further complexity of molecular clusters, gas-liquid and solid-liquid surfaces, reactions between these materials, the consequences of physical and electromagnetic processing and the addition of ethanol are considered. Any or a combination of these factors may cause 'memory' of past solutes and processing in water. Some of these solutions are capable of causing non-specific clinical effects whereas others may cause effects specifically linked to the solution's (and laboratory) history, as outlined below .
Mechanisms for 'the memory of water' as aplied to homeopathy
Specific clinical effects Non-specific clinical effects
Remaining material on surfaces
Aerosol material reintroduced
Bacterial material introduced
Remaining particle clusters Silicates, dissolved and particular
nanobubbles and their material surfaces
Redox molecules produced from water
Natural water clustering
Stabilized water clustering
Ions, including from glassware
Ethanol solution complexity
There are numerous examples of the slow equilibration in aqueous solution. Thus, it can take several days for the effects of the addition of salts to water to finally stop oscillating  and such solutions are still changing after several months showing a large-scale (~100 nm) domain structure . Also, water restructuring after infrared radiation persists for more than a day , and water photoluminescence changes over a period of days . Changes to the structure of water are reported to last for weeks following exposure to resonant RLC (resistance inductance capacitance) circuits . Conductivity oscillations (~ 0.5 Hz) at low concentrations of salts also show the poor tendency to equilibrium in such solutions . Succussion, by itself, has been shown to be 'remembered' for at least 10 minutes as solitons (that is, standing waves) .
Explanation of homeopathy on the basis of water crystals (IE, [124, 125]) is unconvincing as such crystals appear to be artifacts and, even as proposed, the effect of body fluid ions would be to immediately 'dissolve' them.
There is a strange occurrence, similar to the ‘memory of water’ but unconnected to it, in enzyme chemistry where an effectively non-existent material still has a major effect; enzymes prepared in buffers of known pH retain (remember) those specific pH-dependent kinetic properties even when the water is removed such that no hydrogen ions are present ; these ions seemingly having an effect in their absence somewhat against common sense at the simplistic level.b
a A related phenomenon may be the occurrence of conductivity oscillations (~ 0.5 Hz) at similar concentrations of salts at the low concentration limit of obedience to Kohlrauch's law (Onsager's formula) Λm = Λmo - αc½, where Λmo is the limiting molar conductivity, α is a constant and c is the molar concentration .
b This example of pH memory was later explained briefly as the enzymes' acidic and basic groups retaining their charge when in an anhydrous environment . This explanation is accepted but remains unproven independently, is derived from a circular argument and does not inform on how the charge is retained. There remains some puzzle to the extent that a single group in a molecule can either be charged or not charged; it cannot be fractionally charged. Thus the enzyme might be expected to behave as containing a mixture of charged and uncharged groups rather than, as found, fractionally charged groups as in the hydrated enzyme. Perhaps there is sufficient hydration water retained to ensure this, but I do not believe that this has been shown. Whatever, the ‘puzzle’ of the enzyme’s memory disappears with the appearance of an acceptable explanation.