The Placebo Effect Was Discovered by Accident, While Debunking Something Else
Early nineteenth-century English medical dictionaries defined placebo as a medicine given more to please the patient than to benefit him. The word is Latin — I shall please — and it reached medicine as a mild insult, borrowed from the Vespers for the dead, where the line placebo Domino was sung by mourners who had been paid to attend. A placebo was a courtesy. It was what you handed over when you had nothing left to hand over.
That stayed roughly the meaning for a long time: the sugar pill, the entry in the other column of the trial, the nothing you compare the something against. Nobody was in a hurry to notice that the nothing was doing work.
The people who noticed first were not looking for it. They were in Paris, and they had been ordered to investigate a fashionable healer.
What the placebo effect is
The placebo effect is a genuine, measurable change in how a person feels or functions that is produced by the context around a treatment — expectation, ritual, conditioning, the sense of being attended to — rather than by any active ingredient in the treatment itself. The pill has nothing in it. Something still happens. The something is in the patient, and it is not imaginary in the sense of being made up.
One distinction is worth getting straight before anything else, because most of the wilder claims live in the gap. What happens inside a placebo group is the placebo response, and it contains several things at once: people recover from illnesses on their own, symptoms that were unusually bad on the day of enrolment drift back towards their average, and patients report differently to a friendly researcher than to a clipboard. The placebo effect proper is only what remains after you subtract all of that — and you can only subtract it if you have a third arm in the study with no treatment at all. Reviews built on exactly that three-arm design, notably by the Danish researchers Hróbjartsson and Gøtzsche, found the leftover effect much smaller than the folklore, and concentrated almost entirely in one place: symptoms the patient reports, above all pain.
That is a real effect in a narrow lane. Most of what gets called the power of placebo is the wider lane, and it is mostly bookkeeping.
Paris, 1784: the finding nobody had a word for
The royal commission that investigated Franz Mesmer's animal magnetism is normally filed under debunking, and I have told that story from the debunking side in mesmerism. Read the report the other way round and it stops being a refutation and becomes a discovery report.
Benjamin Franklin and Antoine Lavoisier were on that commission. Their method was to separate being treated from believing you are being treated, and then watch which one the symptoms followed. Subjects who were magnetized without their knowledge felt nothing. Subjects who believed a magnetizer was at work behind a door — with nobody there — went into full convulsive crises.
The negative half of that result is famous: no fluid. The positive half is the one nobody had vocabulary for. They had just demonstrated, under blinded conditions, that belief alone reliably produced trembling, weeping, collapse and relief in people who were not pretending. They wrote it up as imagination, which in the idiom of 1784 meant approximately nothing, and the field moved on.
It took medicine roughly a century and a half to build the randomised controlled trial around precisely their logic, and then to work out that the control arm was not empty. The commission that killed animal magnetism had, in the same set of experiments, produced the first controlled demonstration of the placebo effect. Both halves came out of the same afternoon in the same garden. It is the only case I know of where a discipline's founding debunk and one of its founding phenomena arrived in the same envelope.
What the effect is made of
Expectation is the largest single component and the easiest to manipulate. Tell someone a cream is a strong local anaesthetic and their pain ratings drop; tell them the same cream is inert and they do not. Expectation is also physically traceable: placebo analgesia can be reduced by naloxone, a drug that blocks opioid receptors, which means the belief is recruiting the body's own opioid system rather than merely changing what people are willing to say out loud. In Parkinson's disease, placebo responses have been observed alongside dopamine release in the striatum. Whatever is happening, the pharmacology is real even when the pill is not.
Conditioning is the quieter component. Pair a distinctive-tasting drink with an immunosuppressant drug often enough and the drink alone starts to shift immune measures — work that goes back to Robert Ader's laboratory. The body learns associations the way it learns anything else, and a syringe, a white coat and a clinical smell are all cues with long histories attached.
Then there is meaning, the part practitioners are least comfortable discussing. Injections outperform capsules. A clinician who sits down, makes eye contact and explains what will happen produces better outcomes with the same prescription than one who does not. That is the treatment context doing what treatment contexts did for centuries before anyone isolated a molecule.
Nocebo: the same machinery, running backwards
Expectation is not a mood-lifting device. It is a prediction engine, and it works just as well on the downside. Warn a patient about a side effect and the rate of that side effect rises in the group receiving the inactive pill. People drop out of placebo arms because of adverse events they attribute to a substance they never received.
This creates a genuine ethical bind rather than a curiosity. Informed consent requires listing side effects. Listing side effects reliably produces some of them. There is no clean way out, only careful phrasing, and clinicians who think about this seriously spend a lot of time on how a warning is worded rather than whether it is given.
The honest placebos
The standard assumption was that the effect runs on deception: if you know the pill is inert, the trick fails. That assumption has taken serious damage.
Open-label placebo trials give patients pills that are openly, explicitly described as containing no active ingredient. The patients are told, in plain language, that placebos have been shown to produce real effects through conditioning and expectation, and that taking them faithfully matters. Work led from Harvard's placebo research programme, principally by Ted Kaptchuk, has run this design in irritable bowel syndrome, chronic low back pain and cancer-related fatigue, and the honest pills have repeatedly outperformed no treatment.
I would not oversell it. These trials cannot be blinded, they are mostly small, and the ritual of taking a pill twice a day with a rationale attached is itself a powerful frame — the patient has been told to expect something, just told it truthfully. But the deception was supposed to be load-bearing, and it turns out not to be. That is a genuine result, and it changes what the effect can ethically be used for.
Where it stops
The placebo effect is strongest and most reliable in pain, nausea, itch, fatigue, anxiety, sleep quality and mood — everything the patient experiences and reports. It weakens sharply the moment you ask for something a machine can measure and the patient cannot influence.
It does not shrink tumours. It does not clear infections, mend fractures, lower cholesterol or eliminate parasites. A patient given a placebo may honestly feel better while their disease continues on its course unaltered, and that combination has killed people who postponed real treatment because they felt fine. The effect changes the experience of illness far more than it changes the illness. Nothing on this page is medical advice, and no ritual — mine included — substitutes for a doctor.
So is hypnosis just an elaborate placebo?
This is the question I get from the sharper end of the room, and "no, obviously not" is not an answer, it is a flinch.
Start with the concession. A substantial part of what hypnosis produces is expectancy, and Irving Kirsch built a career on saying so. The induction, the practitioner's manner, the framing of the session, the fact that both people have agreed something unusual is about to happen — all of it is placebo machinery in the strict sense, and it is doing real work. Suggestion without any induction at all still produces most of the effect in most people; the formal induction adds a modest increment on top. Any hypnotist who claims their results are free of context effects is describing a practice that does not exist.
Now the part that does not reduce. Hypnotic responsiveness tracks a stable individual trait that behaves like a trait — measurable, roughly normally distributed, durable across years — and attempts to show that placebo responsiveness lives in the same people have not produced a clean result. Naloxone, which reduces placebo analgesia, does not reliably reverse hypnotic analgesia, which points to at least partly different routes to the same relief. And hypnotic suggestion is used as a laboratory instrument to produce specific, targeted phenomena on demand — a limb that will not move, a colour that will not register, a Stroop effect that stops interfering. You do not build a research method on politeness. If you want the longer version of that argument, it is in is hypnosis real, and the mechanism side sits in how hypnosis works.
The fair verdict is that hypnosis is not merely a placebo, and it is not placebo-free either. The two overlap heavily at the level of expectation and diverge at the level of specificity and trait. The clearest place to watch the non-placebo part operate is the ideomotor effect, where an idea moves a hand that its owner is genuinely not moving on purpose — and where the movement stops the moment you arrange things so the person cannot see what they are supposed to expect.
A practitioner who stripped every placebo component out of their work — the calm room, the confident voice, the sense that this will do something — would not be more honest. They would just be worse at it. What matters is that the frame is used with the person's knowledge and consent rather than against it, which is the whole difference between a ritual and a con. If you want to know where you personally sit on the responsive scale before forming an opinion, the suggestibility test settles it faster than reading does, and what two consenting adults can build on that once the arguing stops is what my books are for.
Mesmer never knew what he had found. Neither did the men who took it away from him.
— Mira
FAQ
What is the placebo effect? A genuine change in how someone feels or functions that comes from the context of a treatment — expectation, ritual, conditioning, being cared for — rather than from any active ingredient. It is strongest for symptoms the patient experiences directly, especially pain.
Who discovered the placebo effect? No single person, but the earliest controlled demonstration came from the 1784 royal commission in Paris that investigated Mesmer's animal magnetism, with Benjamin Franklin and Antoine Lavoisier among its members. They showed under blinded conditions that effects followed what subjects believed was being done to them, not what was actually done.
Does a placebo work if you know it is a placebo? Apparently yes, at least sometimes. Open-label trials in conditions like irritable bowel syndrome and chronic low back pain have found that honestly labelled inert pills still beat no treatment, which undercuts the assumption that deception is required.
Is hypnosis just a placebo? Not just, but not separate from it either. Expectation is a large part of how hypnosis works, while hypnotic responsiveness tracks a stable trait that placebo responsiveness does not map onto cleanly, and hypnotic analgesia behaves differently under opioid-blocking drugs than placebo analgesia does.