Homeopathy's rule is simple: dilute the remedy 1:100, shake, repeat — and the more you
dilute, the stronger it gets. Its flagship potency, 30C, means thirty of those
steps. It is the cleanest test case in this whole series, because the placebo response
it rides on is actually real, the honest-placebo question
is genuinely open, and "water remembers" is an
overclaim — and the load-bearing fact is one line of
arithmetic older than the remedy itself: how many molecules are left?
Avogadro published the number that answers it in 1811 — one year after Hahnemann's
Organon. The kill test was ready before the claim finished printing.
Fig. 1 — The dilution line. Top: one mole of active substance through successive 1:100 steps; past the magenta wall the expected molecule count drops below one. Bottom: expected molecules per dose on a log scale — the gold line is a single molecule; your chosen potency is the marker.
Take the honest core first. Serial dilution is real, standard lab technique, and the
arithmetic is exact. Start with a full mole of active substance — Avogadro's number,
$N_A \approx 6.022\times10^{23}$ molecules — and each centesimal step keeps one part
in a hundred:
That crossing is the Avogadro wall. Below 12C a remedy can still
contain what its label claims — at 6C there are a healthy $6\times10^{11}$ molecules
in the dose. At the wall, the expectation falls below one molecule, and past
it the dilution is arithmetic theater: at 30C, the flagship potency
sold in every pharmacy, the chance your dose contains a single molecule of medicine
is about $6\times10^{-37}$. To meet one molecule you would need
$1.7\times10^{36}$ doses — a sphere of remedy ten and a half solar radii
across, or about $10^{12}$ Earth-oceans.
And 30C is the modest option. Oscillococcinum, a top-selling flu remedy, is
200C: $N \approx 10^{-376}$, against roughly $10^{80}$ atoms in the observable
universe. There is no possible universe of doses large enough to contain what the
label describes.
Slide the potency and watch both panels. The point is not that the remedy is
poisoned or fake — it's water, prepared with great care. The point is that
past the wall, whatever helps a patient cannot be chemistry, which hands the entire
question to the one layer that survives: the placebo response, which is real,
measurable, and doesn't need the vial.
The same claim, read by parallax — from the measured placebo to the memory of water.
actually real
The placebo response is real medicine-adjacent science.
Signal. Placebo effects, regression to the mean, and the measurable value of a long, attentive consultation are genuine, quantified clinical phenomena — for pain, nausea, and other self-reported outcomes, robustly so. A homeopath's hour of undivided attention is not nothing; it's just not in the vial.
Kill test. Placebo-controlled trials exist precisely to measure this layer, and they measure it every day. It survives — as the explanation, not the remedy.
genuinely open
Can a placebo work when you know it's a placebo?
Signal. Open-label placebo trials (Kaptchuk and successors, 2010–) report benefit for IBS, chronic pain, and fatigue even when patients are told, plainly, that the pill is inert — no deception required.
Unresolved. Effect sizes, durability, and which conditions respond are genuinely undecided; the trials are small and the outcomes self-reported. This is the live, ethical version of the question homeopathy accidentally raises.
Kill test. Larger pre-registered open-label trials with objective endpoints — running now. A real question with a real test still in progress.
real kernel · overreached
"Water remembers what it once contained."
Signal. The kernel is a real episode of science: Benveniste's basophil results were produced in a real lab and published in Nature (1988) — taking the claim seriously enough to test was the system working.
Noise. Under blinded re-runs supervised by Maddox, Randi and Stewart, the effect vanished. And the physics allows no refuge: water's hydrogen-bond network reshuffles in picoseconds, while a remedy on a shelf would need its "memory" to persist about $3\times10^{19}$ times longer.
Kill test. Blind the samples. It was run, in Benveniste's own lab, in 1988. The memory did not survive the blindfold.
Signal. The mood is honest: gentler medicine, more listening, fewer side effects. In the 1810s, when "heroic" medicine meant bloodletting and mercury, doing chemically nothing was genuinely safer — homeopathy's early success was real, for that reason.
Noise. Past the Avogadro wall there is nothing present to potentiate, no dose–response curve, and no proposed carrier that survives contact with the picosecond physics of water. Succussion adds kinetic energy, then the glass forgets.
Kill test. Aggregate the best trials: Shang et al. (Lancet, 2005) and Australia's NHMRC review (2015) — effects compatible with placebo across conditions. The claim persists only by forbidding no result.
Notice the sift: the real layer (placebo, attention, regression to the mean) explains
everything the believers actually experience; the open layer asks the interesting
modern question; and the overclaim was dead on arrival — Avogadro ran the kill
test two centuries in advance, one year after the remedy was invented.