The Placebo Effect: What It Actually Changes — and What It Never Does

Three numbers worth starting with

Each comes from a separate randomised trial, all linked at the end. Together they describe the placebo effect more honestly than any definition.

A word that started life as an insult

“Placebo” is Latin for “I shall please”. It reached medicine not from a laboratory but from a church service: the Vespers for the Dead opened with a line from the Psalms — “Placebo Domino in regione vivorum”, “I shall please the Lord in the land of the living”. Professional mourners were hired to sing that line at funerals of people they had never met. By the fourteenth century “placebo” meant a flatterer — Chaucer named a sycophantic character exactly that in The Merchant's Tale.

The word entered medical vocabulary carrying the same sneer. Hooper's medical dictionary of 1811 defines a placebo as “an epithet given to any medicine adapted more to please than benefit the patient”. It was professional slang with contempt built in: something handed over to end the conversation.

What happened next is the interesting part. The thing dismissed for two centuries as polite emptiness turned out, in the second half of the twentieth century, to be a measurable biological phenomenon — and what began as a jab at doctors became a research field with its own laboratories at Harvard and in Turin.

Three experiments that settled it: this is physiology, not politeness

The obvious objection to the placebo effect sounds reasonable: the patient simply says they feel better because that's what's expected. These three studies close that objection from different directions — by measuring the chemistry of the brain rather than the words of the patient.

Placebo can be trained, the way Pavlov trained a dog

Expectation is not the only mechanism. The second is conditioning, and it was found by accident. In 1975 the psychologist Robert Ader and the immunologist Nicholas Cohen gave rats saccharin-sweetened water together with cyclophosphamide, a drug that suppresses the immune system and causes nausea. The rats duly developed a taste aversion to saccharin. Then came the part nobody planned: when the animals were given the sweetened water alone, with no drug at all, their antibody production fell. The body had learned to suppress its own immune response in reaction to a taste.

To check that this was not simply stress from an unpleasant flavour, the authors repeated the experiment with lithium chloride — a substance that produces the same taste aversion but no immune suppression. That group showed no immune response. The body had learned not “something bad”, but the specific pharmacological action of a specific drug.

In 2002 Manfred Schedlowski's group demonstrated the same thing in people. Healthy volunteers took cyclosporin A — a powerful immunosuppressant used after organ transplants — alongside a distinctively flavoured drink for several days. A week later they were given the same drink and placebo capsules instead of the drug. Their immune function measurably dropped: reduced IL-2 and interferon-gamma expression and reduced lymphocyte proliferation. These were not patient reports; they were numbers from a laboratory.

The practical implication is not that we should treat illness with flavoured drinks. It is that placebo response is partly learned: the more times in your life you have felt better after a white pill, the more reliably the next white pill works. The effect accumulates with experience — which is why it runs strongest in cultures with high trust in the pharmacy.

The most elegant experiment in the field: asthma

In 2011 a team at Harvard and Brigham and Women's Hospital did something researchers rarely bother with: they added a fourth arm to the usual drug-versus-placebo comparison — “do nothing at all”. Thirty-nine asthma patients cycled through all four conditions across twelve visits each. At every visit two different things were measured: how much air the patient could actually force out (FEV₁ — an instrument, not an opinion) and how much better they felt.

Wechsler et al., New England Journal of Medicine, 2011. The two bars in each pair are the same person in the same trial. By self-report, the dummy inhaler nearly matched the drug (45–46% versus 50%). By lung function, the real inhaler produced a 20% increase and both placebos produced what doing nothing produced. Placebo changed how the patient felt and left the airways exactly as they were.

The line popular science tends not to draw

The asthma trial is not a curiosity — it illustrates the most rigorous work in the field. In 2001 the Danish researchers Asbjørn Hróbjartsson and Peter Gøtzsche did what nobody had done before: they gathered 130 trials in which placebo was compared not against a drug but against no treatment at all — that is, against the natural course of the illness.

The result was far more modest than the legend. On binary outcomes (recovered or not, heart attack or none), placebo produced no statistically significant effect whatsoever. On continuous measures there was an effect — but essentially only where the measure was subjective, meaning the patient reported it. On objective measured outcomes, no significant effect.

The single most useful figure from that work concerns pain. Placebo did reduce pain compared with nothing — by an average of 6.5 millimetres on a 100-millimetre scale. That is a real effect. It is not a miracle. The authors also noticed the effect shrank in larger trials — the classic signature of small, weaker studies inflating an average.

And there is one more explanation regularly mistaken for placebo: regression to the mean. People don't seek help on a random day, they seek it on their worst one. Things then almost always improve somewhat, purely because the worst day was a peak rather than a new baseline. If they happened to take something at that moment, that is what gets the credit. Without a group that took nothing, the two are impossible to tell apart — which is exactly why the asthma trial's fourth arm is so valuable.

The one sentence this article exists for

Nocebo: the same mechanism running backwards

In 2020 cardiologists at Imperial College London took 60 patients who had previously abandoned statins because of muscle pain and gave each of them a year made of twelve monthly periods in random order: four months on a statin, four on placebo, four on no tablet at all. Patients rated symptom intensity daily from 0 to 100.

SAMSON, Wood et al., New England Journal of Medicine, 2020. The gap between statin and placebo was not statistically significant. The gap between “any tablet” and “no tablet” was nearly twofold. The authors calculated a nocebo ratio: 90% of the symptom burden patients felt on the statin, they also felt on the dummy. The symptoms were entirely real — their source was the act of taking a pill, not the drug in it. Six months later, half the participants had gone back on statins.

How big this gets outside a trial

The largest natural experiment in nocebo ever run was the COVID-19 vaccine programme. A 2022 meta-analysis pooled 45,380 participants from twelve trials, of whom 22,578 received a placebo injection — saline.

After the first dose, 35.2% of those saline recipients reported systemic adverse events, most commonly headache (19.3%) and fatigue (16.7%). Comparing that with the vaccine arms, the authors calculated that nocebo responses accounted for 76% of all systemic adverse events after the first dose, and around 52% after the second.

This does not mean vaccine side effects are imaginary — after the second dose the gap between vaccine and placebo was large and entirely real. It means something else: a substantial share of what people genuinely felt and honestly described came from expectation rather than from the contents of the vial. For medical communication that matters more than it sounds: reading out a detailed list of possible side effects before a procedure measurably raises the odds that the patient will experience them.

What amplifies the effect — none of it pharmacological

Placebo response is not a constant. Dozens of studies have shown it shifts with factors that have nothing to do with any active ingredient, because the brain is grading not the chemistry but the apparent seriousness of the intervention.

It still works when nobody is being deceived

The most counterintuitive result of the last two decades: the effect does not vanish when the secrecy does. In 2010 Ted Kaptchuk handed patients with irritable bowel syndrome bottles labelled “placebo” and told them plainly that these were inert pills with no active medication, but that they might produce significant relief through mind-body self-healing processes. The patients reported statistically significant improvement compared with a group given nothing.

A 2021 meta-analysis pooled eleven such trials — back pain, cancer-related fatigue, ADHD, allergic rhinitis, depression, IBS, menopausal hot flushes — and found a moderate overall effect (standardised mean difference 0.72).

Here honesty is required, and it is usually missing from retellings. The same authors state plainly that the trials are few, heterogeneity is high (I² = 76%), risk of bias is moderate, and the field is “in its infancy”. Open-label placebo is a promising scientific finding, not a ready-made treatment. And like every placebo, it acts on how people feel, not on the disease.

Why some people say “everything works on me” and others say nothing does

Attempts to find a “placebo personality” failed: no stable psychological profile of a placebo responder has held up. What did turn up is more interesting — genetics. In a 2012 study, IBS patients were sorted by variant of the COMT gene, which encodes the enzyme that breaks down dopamine. Carriers of the met/met variant — slower dopamine breakdown, therefore higher levels — responded to placebo treatment markedly more strongly than the rest.

That launched a research direction called the “placebome”, by analogy with the genome: the search for genetic markers of placebo response. Which means part of the everyday observation that “everything helps me” or “nothing touches me” may have a concrete biochemical cause — and it isn't about the drug.

And now the most useful fact in this article for a sceptic. In 2010 veterinary researchers pooled three placebo-controlled trials of canine epilepsy: 22 of 28 dogs (79%) on placebo had fewer seizures, and 8 of them (29%) formally qualified as responders with at least a halving of seizure frequency. A dog does not know it is being treated and holds no expectations. But the owner counting the seizures does know — and counts differently. Part of what we habitually call the placebo effect is an observer effect: what changes is not the illness but the recording of it.

Placebo is getting stronger — and it is breaking drug development

In 2002 an analysis of major depression trials showed that the proportion of patients “helped” by a dummy pill had risen statistically from the 1980s through the 2000s. Depression had not become milder; the context had — drug advertising, awareness, the expectations patients bring to treatment.

In 2015 Jeffrey Mogil's team at McGill turned that observation into something far more precise. They analysed trials of drugs for neuropathic pain from 1990 to 2013 and found that placebo responses had risen substantially while responses to the drugs themselves stayed flat. The margin of drug over dummy was narrowing — and not because the drugs got worse.

The strangest detail: this rise happened only in trials conducted in the United States. Elsewhere, placebo response was unchanged. The authors link it to the fact that American trials grew larger and longer over those years, and both size and duration are statistically associated with a stronger placebo response. The US is also one of only two countries in the world — with New Zealand — that permits direct-to-consumer advertising of prescription drugs.

The practical consequence is unpleasant: some potentially useful molecules fail registration trials not because they do nothing, but because their competitor in the control arm got stronger.

Seventy years: from legend to measurement

The history of placebo research is not a story of discovery but of correction: the effect was first overstated, then the overstatement was demolished, then the mechanism was found and its limits established.

1. 1955

   Henry Beecher publishes “The Powerful Placebo”

   The anaesthetist calculates that an average of 35% of patients respond to a dummy. The figure launches the fashion for placebo-controlled trials — and still circulates in popular writing today.

2. 1975

   Ader and Cohen: the immune system can be trained

   Rats suppress their own immune response to the taste of sweetened water. Placebo turns out to be not only expectation but a learned reflex.

3. 1978

   Naloxone switches off placebo pain relief

   The first direct evidence that placebo analgesia runs on the brain's own opioid system rather than on patient courtesy.

4. 1997

   Kienle and Kiene re-examine Beecher — and find no effect

   Re-analysis of the same 15 studies: most “improvements” are explained by the natural course of illness and regression to the mean. The famous 35% turns out to be an artefact.

5. 2001

   Two opposite blows in one year

   PET shows genuine dopamine release from a dummy in Parkinson's. The same year, a meta-analysis of 130 trials shows placebo does not move objective outcomes. Both results are correct — they are simply about different things.

6. 2010

   Placebo without deception

   Patients explicitly told they are receiving a dummy still improve. The ethical dead end partly dissolves.

7. 2011

   Asthma puts everything in its place

   A dummy matches the drug on how patients feel and does nothing to their lung function. The cleanest demonstration of the effect's boundary.

8. 2020

   SAMSON: side effects come from expectation too

   90% of “statin side effects” reproduce on a dummy pill. Nocebo moves from theory into cardiology practice.

9. 2022

   COVID vaccines measure nocebo across 45,000 people

   76% of systemic side effects after the first dose also appeared in the saline group.

Four myths worth discarding

What to do about it: the ritual can be given honestly, the pill often can't

Placebo left the laboratory long ago. A 2008 survey of 679 American internists and rheumatologists found that roughly half of them prescribe “placebo treatments” on a regular basis. Sugar pills: almost nobody (2%). Saline: 3%. What they actually use are over-the-counter analgesics (41%) and vitamins (38%). A further 13% admitted to prescribing antibiotics in that role, and the same share sedatives. Patients are rarely told: only 5% describe the prescription as a placebo.

That is where the real ethical line runs — not between honest and dishonest doctors, but between a ritual and a product. The ritual of care — time, attention, regular follow-up, the sense of being heard — can be given honestly and for free; the research on the components of placebo says so directly. Selling a product with no evidence of efficacy at retail price is something else: it monetises the trust people place in the word “medicine” on a box.

And the line is not merely philosophical. A meta-analysis of the full dataset submitted by manufacturers to the US Food and Drug Administration — including unpublished negative trials — showed that the clinically meaningful gap between antidepressant and placebo appears reliably only in severe depression, while in mild and moderate cases most of the effect is statistically indistinguishable from a dummy. That is not a reason to stop prescribed treatment; it is a reason to understand that the boundary between “a drug” and “a placebo reinforced by packaging” is thinner than the marketing suggests.

The biggest practical risk is not that someone “feels better” from a cold remedy. It lies in three places: in the thirteen per cent of antibiotics prescribed as placebos, and the resistance that breeds; in money spent on an effect available for free; and in time — when a product that works through expectation delays treatment in a situation where timing decides the outcome. Placebo is excellent at relieving discomfort. It does not treat pneumonia.

Seven terms that change how you read any trial

Placebo

An intervention with no specific active ingredient — a pill, an injection, a procedure. The placebo effect is the change in state after such an intervention compared with no treatment at all.

Nocebo

The same mechanism in reverse: negative expectation producing real unpleasant symptoms. The nocebo ratio expresses what share of a drug's side effects reproduce on a dummy.

Open-label placebo

A placebo the patient is explicitly told is a placebo. Weaker than the deceptive version but stronger than nothing — and it requires no deception.

Regression to the mean

People seek help on their worst day. Things then usually improve regardless of treatment. The single most common reason recovery gets mistaken for a drug working.

Objective endpoint

What an instrument or laboratory measures: forced expiratory volume, tumour size, cholesterol, mortality. Placebo barely moves these.

Subjective endpoint

What the patient reports: pain, nausea, fatigue, anxiety. This is where the placebo effect is measurable and real.

Double blinding

Neither patient nor clinician knows who is getting the drug and who the dummy. Without it, the expectations of both sides leak into the result and the trial stops proving anything.

Common questions

Sources

Every study mentioned in the text is a primary publication indexed in PubMed, the US National Library of Medicine database. Each link opens the original.

1. de la Fuente-Fernández R. et al. Expectation and dopamine release: mechanism of the placebo effect in Parkinson's disease. Science, 2001.
2. Levine J.D., Gordon N.C., Fields H.L. The mechanism of placebo analgesia. Lancet, 1978.
3. Amanzio M., Pollo A., Maggi G., Benedetti F. Response variability to analgesics: a role for non-specific activation of endogenous opioids. Pain, 2001.
4. Colloca L., Lopiano L., Lanotte M., Benedetti F. Overt versus covert treatment for pain, anxiety, and Parkinson's disease. Lancet Neurology, 2004.
5. Ader R., Cohen N. Behaviorally conditioned immunosuppression. Psychosomatic Medicine, 1975.
6. Goebel M.U. et al. Behavioral conditioning of immunosuppression is possible in humans. FASEB Journal, 2002.
7. Wechsler M.E. et al. Active albuterol or placebo, sham acupuncture, or no intervention in asthma. New England Journal of Medicine, 2011.
8. Hróbjartsson A., Gøtzsche P.C. Is the placebo powerless? An analysis of clinical trials comparing placebo with no treatment. New England Journal of Medicine, 2001.
9. Wood F.A. et al. N-of-1 Trial of a Statin, Placebo, or No Treatment to Assess Side Effects (SAMSON). New England Journal of Medicine, 2020.
10. Haas J.W. et al. Frequency of Adverse Events in the Placebo Arms of COVID-19 Vaccine Trials: A Systematic Review and Meta-analysis. JAMA Network Open, 2022.
11. Waber R.L., Shiv B., Carmon Z., Ariely D. Commercial features of placebo and therapeutic efficacy. JAMA, 2008.
12. Blackwell B., Bloomfield S.S., Buncher C.R. Demonstration to medical students of placebo responses and non-drug factors. Lancet, 1972.
13. Moseley J.B. et al. A controlled trial of arthroscopic surgery for osteoarthritis of the knee. New England Journal of Medicine, 2002.
14. Kallmes D.F. et al. A randomized trial of vertebroplasty for osteoporotic spinal fractures. New England Journal of Medicine, 2009.
15. Buchbinder R. et al. A randomized trial of vertebroplasty for painful osteoporotic vertebral fractures. New England Journal of Medicine, 2009.
16. Kaptchuk T.J. et al. Components of placebo effect: randomised controlled trial in patients with irritable bowel syndrome. BMJ, 2008.
17. Kaptchuk T.J. et al. Placebos without deception: a randomized controlled trial in irritable bowel syndrome. PLoS One, 2010.
18. von Wernsdorff M. et al. Effects of open-label placebos in clinical trials: a systematic review and meta-analysis. Scientific Reports, 2021.
19. Hall K.T. et al. Catechol-O-methyltransferase val158met polymorphism predicts placebo effect in irritable bowel syndrome. PLoS One, 2012.
20. Muñana K.R., Zhang D., Patterson E.E. Placebo effect in canine epilepsy trials. Journal of Veterinary Internal Medicine, 2010.
21. Walsh B.T. et al. Placebo response in studies of major depression: variable, substantial, and growing. JAMA, 2002.
22. Tuttle A.H. et al. Increasing placebo responses over time in U.S. clinical trials of neuropathic pain. Pain, 2015.
23. Beecher H.K. The Powerful Placebo. JAMA, 1955.
24. Kienle G.S., Kiene H. The powerful placebo effect: fact or fiction? Journal of Clinical Epidemiology, 1997.
25. Tilburt J.C. et al. Prescribing “placebo treatments”: results of national survey of US internists and rheumatologists. BMJ, 2008.
26. Kirsch I. et al. Initial severity and antidepressant benefits: a meta-analysis of data submitted to the Food and Drug Administration. PLoS Medicine, 2008.