Give someone an inert pill they believe is a painkiller and their pain often falls, their brain releases its own opioids, and the relief can be blocked by an opioid-blocking drug. That is the placebo effect. It is the body responding to expectation, learning, and the ritual of being treated, and it is neither imaginary nor fake.
The care is all in the boundary: placebos are strong for how a person feels and reports, and they do not shrink a tumor, mend a fracture, or clear an infection. Treating a serious disease with belief in place of effective care is where the harm lies, and holding that line is also why this section weighs controlled trials over stories.
What It Is
A placebo is an inert treatment given in a setting that leads a person to expect benefit: a sugar pill, a saline injection, a sham procedure. The placebo effect is what the body does and what the person reports as a result. Three things drive it.
- Expectancy is the conscious belief that relief is coming.
- Conditioning is associative learning. A body that has repeatedly felt relief after taking a pill or entering a clinic comes to respond to the pill and the clinic themselves, before any active drug acts.
- The therapeutic ritual is the whole encounter: the attention, the explanation, the authority and warmth of the person treating you.
None of this means the symptom is invented. Placebo pain relief runs on the body's own opioid system, and in Parkinson's disease the expectation of a drug releases measurable dopamine. These are physiological events, described in the section below.
It Moves Symptoms, Not Disease
Most accounts blur this distinction. Placebos act powerfully on subjective, self-reported symptoms, pain, nausea, and fatigue, and much less, if at all, on the objective course of disease.
The cleanest evidence comes from two large syntheses that gathered every trial comparing a placebo against no treatment at all. Across more than 200 trials, placebos produced a modest effect on patient-reported continuous outcomes, clearest for pain, and no meaningful effect on binary or objectively measured ones.
A placebo can make a person hurt less. It does not shrink a tumor, mend a fracture, or clear an infection. Treating a serious disease with belief in place of effective care is where the harm lies: someone delaying treatment for a condition a placebo was never going to touch.
When a cancer appears to regress or a chronic illness to resolve after an inert treatment, the explanation is the natural ups and downs of the illness, not the sugar pill. The placebo effect is strong exactly where symptoms are felt and reported, and small to absent where a scan or a blood test does the measuring. The claim that belief can cure serious disease takes a documented effect and stretches it past where the data reach.
The Research & Studies
Everything here is based on the research we have collected and checked, sorted into groups and ordered with the strongest evidence first. Click any claim to open the studies behind it.
How it works
Naloxone reverses placebo pain relief, so the body's own opioids do the work
People who got real pain relief from a placebo after dental surgery lost that relief when given a drug that blocks opioids. The placebo relief was running on the body's own opioids, not on imagination.
Levine, Gordon and Fields studied pain following extraction of impacted third molars. Patients who responded to a placebo with reduced pain lost that analgesia when given naloxone, an opioid antagonist, whereas naloxone had less effect where placebo had not produced relief. The reversal established that at least one major route of placebo analgesia is the release of endogenous opioids, giving the placebo effect a concrete neurochemical mechanism. Later imaging and pharmacological work has repeatedly confirmed opioid involvement in expectation-driven analgesia. Sex distribution of the dental-surgery sample was not reported in a way that supports a breakdown.
The study · 1
Levine, Gordon & Fields, The mechanism of placebo analgesia · Lancet 1978;2(8091):654-657
Placebo eases self-reported symptoms, mainly pain, across 202 trials
When researchers pooled every trial that pitted a placebo against getting nothing, the placebo made people feel modestly better on things they report themselves, above all pain. It did not change outcomes measured by a scan or a test.
The 2010 Cochrane review by Hrobjartsson and Gotzsche pooled 202 randomized trials in which one arm received a placebo and another received no treatment. Placebo showed a modest overall effect on continuous, patient-reported outcomes, most reliably pain, with the size varying widely by condition and trial quality, and little to no effect on binary outcomes or objective measures. Effects were larger where patients were informed the intervention was intended to strengthen the response and where the outcome was assessed by the patient rather than an instrument. The authors concluded placebo can influence patient-reported outcomes, especially pain and nausea, but the evidence does not support a general clinical placebo effect on the course of disease.
The study · 1
Hrobjartsson & Gotzsche, Placebo interventions for all clinical conditions · Cochrane Database Syst Rev 2010;(1):CD003974
Placebo does not change objectively measured disease across 114 trials
Pooling 114 trials of placebo versus no treatment, the placebo did not change outcomes measured objectively, like lab values or physical findings. Whatever effect there was showed up only in symptoms people rated themselves, mainly pain.
The 2001 New England Journal of Medicine analysis by Hrobjartsson and Gotzsche, Is the placebo powerless?, compared placebo against no-treatment arms in 114 randomized trials. It found no statistically significant effect of placebo on binary outcomes and no important effect on objective continuous outcomes, with a possible small effect limited to continuous subjective outcomes, particularly pain. The finding directly counters the popular idea of a broad, disease-altering placebo effect: placebos do not shrink tumors, mend fractures or clear infections. Apparent recoveries after inert treatment are better explained by the natural course of illness and regression to the mean than by the placebo itself.
The study · 1
Hrobjartsson & Gotzsche, Is the placebo powerless? · N Engl J Med 2001;344(21):1594-1602
Expecting Parkinson's medication releases dopamine like a levodopa dose
When people with Parkinson's expected to get their medication, a brain scan showed their brains released dopamine, the chemical the disease lacks, just from the expectation. The placebo response had a visible chemical signature.
de la Fuente-Fernandez and colleagues used PET with raclopride, a tracer sensitive to dopamine release, in patients with Parkinson's disease. Expectation of receiving antiparkinsonian medication produced substantial release of endogenous dopamine in the dorsal and ventral striatum, comparable in magnitude to a therapeutic dose of levodopa or apomorphine. The result located the placebo response in Parkinson's within the reward and motor circuitry that the disease depletes, showing that expectation of benefit engages the same dopaminergic system the drug targets. It is a mechanistic imaging finding in a specific disease, not a claim that placebos slow Parkinson's progression.
The study · 1
de la Fuente-Fernandez et al., Expectation and dopamine release · Science 2001;293(5532):1164-1166
Placebo lowers activity in the brain's pain-processing regions on fMRI
Brain scans during placebo pain relief showed less activity in the regions that create the feeling of pain, and more activity in planning regions while people anticipated relief. Expectation was reshaping how the brain handled the pain signal.
Wager and colleagues used functional MRI during a placebo analgesia paradigm. Placebo reduced pain-related neural activity in regions including the thalamus, insula and anterior cingulate cortex during noxious stimulation, and increased activity in the prefrontal cortex during the anticipation of pain, consistent with a top-down expectation signal modulating pain processing. The study helped establish that placebo analgesia is not merely a change in reporting but a change in the central processing of the pain signal itself. It maps the expectancy mechanism rather than testing a treatment.
The study · 1
Wager et al., Placebo-induced changes in FMRI in the anticipation and experience of pain · Science 2004;303(5661):1162-1167
A warm practitioner relationship lifted IBS relief from 28% to 62%
IBS patients were split into a waiting list, a brief sham treatment with little interaction, and the same sham given by a warm, attentive practitioner. Reported relief climbed with each step, from about 28% to 44% to 62%. The relationship itself raised how much relief people reported.
Kaptchuk and colleagues randomized 262 patients with irritable bowel syndrome to a waiting list, sham acupuncture with a limited practitioner interaction, or sham acupuncture delivered with an augmented, warm and attentive patient-practitioner relationship. Adequate relief at three weeks rose across the arms, approximately 28% on the waiting list, 44% with limited interaction, and 62% with the augmented relationship, a graded dose-response. The trial isolated the components of the placebo response and showed that the therapeutic ritual and the quality of the clinical relationship contribute an active, measurable effect over and above assessment and natural history. IBS cohorts skew female; the trial enrolled both sexes.
The study · 1
Kaptchuk et al., Components of placebo effect: randomised controlled trial in IBS · BMJ 2008;336(7651):999-1003
90% of statin muscle symptoms also appeared on an identical placebo
People who had quit statins over muscle symptoms took the statin, an identical dummy pill, and nothing, in random order over months. Ninety percent of the symptoms they logged on the statin also appeared on the dummy pill. The symptoms were real, but the drug was not causing most of them.
The SAMSON n-of-1 trial by Wood, Howard and colleagues enrolled 60 patients who had abandoned statin therapy because of adverse symptoms. Each undertook a randomized sequence of months on atorvastatin, months on identical placebo, and months with no tablet, recording daily symptom intensity. The symptom burden during placebo months was 90% of that during statin months, and there was no significant difference in symptom intensity between statin and placebo. This quantifies the nocebo component of statin intolerance: the symptoms are experienced, but most of the intensity is attributable to the act of taking a tablet and the expectation of harm rather than to the drug. Both sexes were enrolled.
The studies · 2
Wood, Howard et al., N-of-1 trial of a statin, placebo, or no treatment to assess side effects (SAMSON) · N Engl J Med 2020;383(22):2182-2184
Colloca & Barsky, Placebo and nocebo effects · N Engl J Med 2020;382(6):554-561
Warning men about finasteride's sexual side effects raised them from 15% to 44%
Men taking finasteride for an enlarged prostate were either told about possible sexual side effects or not. Those who were warned reported these side effects about three times as often, roughly 44% versus 15%, on the identical medication. The warning itself raised the harm.
Mondaini and colleagues studied 120 men taking finasteride for benign prostatic enlargement, randomized as to whether they were informed that the drug could cause erectile dysfunction, reduced libido and ejaculation problems. Men counseled about these effects reported them at 43.6%, against 15.3% among men not informed, for the same medication and dose. The threefold difference is a clear nocebo effect: expectation of a specific harm substantially increased its reported occurrence. Because the trial enrolled only men, taking finasteride for a male condition, the size of any analogous effect in women is untested.
Who this may not transfer to:Measured only in men taking a drug for a male condition. The nocebo mechanism, expectation of harm increasing reported harm, is general and well documented in both sexes elsewhere, but the specific magnitude seen here has not been measured in women and should not be assumed to be the same.
The study · 1
Mondaini et al., Finasteride 5 mg and sexual side effects: how many are a nocebo phenomenon · J Sex Med 2007;4(6):1708-1712
Openly labeled placebo beat no treatment in IBS, 59% versus 35%
IBS patients who were openly told their pills were sugar pills, and that placebos can still help, reported more improvement than patients given nothing, about 59% against 35%. Knowing it was a placebo did not switch off the effect.
Kaptchuk and colleagues randomized 80 patients with irritable bowel syndrome to open-label placebo, pills openly described as inert together with an explanation that placebos can produce benefit through mind-body self-healing processes, or to no treatment. The open-label placebo group reported significantly higher rates of adequate relief and global improvement, on the order of 59% versus 35%, over three weeks. The result challenged the assumption that concealment is necessary for a placebo to work and opened the open-label placebo research line. It is a small, short trial in a symptom-defined condition, and open-label placebos have not been shown to affect objective disease.
The study · 1
Kaptchuk et al., Placebos without deception: a randomized controlled trial in IBS · PLoS One 2010;5(12):e15591
Pooled open-label placebo trials show a large benefit over no treatment (SMD about 0.88)
Pooling the handful of trials where people knowingly took a placebo, there was a sizeable benefit over getting nothing, in conditions like IBS, depression and back pain. The evidence base is still thin.
Charlesworth and colleagues systematically reviewed and meta-analyzed randomized trials of open-label placebo, placebos given with full disclosure, against no treatment. Across a small set of trials in conditions such as irritable bowel syndrome, depression, allergic rhinitis, chronic low back pain and ADHD, open-label placebo showed a large pooled effect (standardized mean difference about 0.88, 95% CI 0.62 to 1.14) on self-reported outcomes. The authors emphasized the small number of trials, modest sample sizes and heterogeneity, and that outcomes were subjective. The synthesis supports open-label placebo as a researchable effect while keeping it firmly at an emerging tier, and confined to symptoms rather than objective disease.
The study · 1
Charlesworth et al., Effects of placebos without deception compared with no treatment · J Evid Based Med 2017;10(2):97-107
The Machinery: Expectation And Conditioning
The reason placebos move some things and not others follows from how they work. Expectation is a signal the brain acts on, and it engages the body's own regulatory chemistry.
Placebo pain relief runs on endogenous opioids. In a study of pain after dental surgery, the relief people got from a placebo was reversed by naloxone, the drug that blocks opioids, which means the body's own opioids were doing the work. Brain imaging during placebo analgesia shows reduced activity in the regions that build the experience of pain, the thalamus, insula, and anterior cingulate cortex, and increased prefrontal activity while a person anticipates relief. In Parkinson's disease, the expectation of an active drug releases measurable dopamine in the striatum, comparable in size to a therapeutic dose of levodopa.
These systems raise or lower the intensity of a symptom. They do not rebuild tissue or kill a pathogen, which is why the effect stops where it does.
The ritual itself has a dose. In a trial in irritable bowel syndrome, 262 patients were split into a waiting list, a brief sham treatment with little interaction, and the same sham delivered inside a warm, attentive, unhurried relationship. Adequate relief rose step by step across the three, from about 28% to 44% to 62%, largest with the attentive relationship. The encounter was an active ingredient with a measurable size.
Open-Label Placebos
Some placebos still work even when the person is told the pill is a placebo. In irritable bowel syndrome, patients openly given pills described as inert, with the explanation that placebos can help through mind-body processes, reported more improvement than patients given nothing, about 59% versus 35%.
Pooled across the handful of trials done so far, in conditions like IBS, chronic low back pain, depression, and fatigue, open-label placebos show a large effect by the usual statistical measure, a standardized mean difference around 0.88 (95% CI 0.62 to 1.14). That size rests on few, small, short trials of self-reported symptoms, so it is not yet settled, and like concealed placebos it does not reach objective disease. What it suggests is that concealment was never the necessary part. The ritual, the expectation, and the act of doing something may carry much of the effect on their own.
The Nocebo Effect
Expectation runs in both directions. When a person expects harm, they can experience it, and this is the nocebo effect. It is not malingering, and it drives a large share of the side effects attributed to drugs.
In the SAMSON trial, 60 people who had abandoned statins because of muscle symptoms each took months of the statin, months of an identical placebo, and months of nothing, in a randomized order. 90% of the symptom burden they recorded on the statin was also present on the placebo, with no significant difference in intensity between the two. The symptoms were felt; most of them were not coming from the drug.
In men taking finasteride for an enlarged prostate, those explicitly warned about sexual side effects reported them about three times as often, 44% versus 15%, for the same medication.
Warning a person about a harm can bring the harm on. This is the same expectancy machinery pointed the other way, and it is why a side-effect warning has to be worded with care.
Why This Means We Trust Controlled Trials
The method follows from the biology. If expectation, conditioning, and the ritual of care can move a symptom on their own, then a person improving after any treatment tells you almost nothing about whether the treatment did the work. Several forces push a symptom toward better whatever you do:
- the placebo response itself,
- the natural course of an illness that was going to ease anyway,
- regression to the mean, the tendency for people to seek help when they feel worst and then drift back toward their average.
A single before-and-after story captures all of that at once and cannot separate it from the treatment. The only way to isolate what a treatment specifically adds is to compare it against a placebo given inside the same ritual, so that expectation and natural history fall on both sides and cancel. What remains is the specific effect, if there is one.
This is why the evidence tables on this site grade findings against controls and treat uncontrolled stories as the weakest signal. It follows directly from the placebo effect being a documented force: because expectation moves symptoms, only a design that holds expectation constant can show what a treatment is worth. The page on how we grade evidence sets out the tiers that follow, and the stress axis page describes some of the physiology that expectation and calm act through.
Common Questions
Is the placebo effect all in my head?
No, and that phrase gets it backwards. The changes are physical. Placebo pain relief releases the body's own opioids and can be blocked by an opioid-blocking drug, and in Parkinson's disease the expectation of medication releases measurable dopamine. What is true is that the effect works on how a symptom is generated and felt, so it is strongest for things like pain, nausea, and fatigue, and it does not reach the underlying disease.
Can a placebo cure a disease or shrink a tumor?
No. This is the boundary that matters most. When trials compared placebos against no treatment at all, placebos moved subjective, self-reported symptoms, clearest for pain, and did not change objectively measured disease. A placebo does not shrink a tumor, mend a fracture, or clear an infection. Apparent cures after inert treatment reflect the natural course of the illness, and relying on belief in place of treatment for a serious disease is where harm happens.
Do placebos work if I know it is a placebo?
Sometimes, for symptom-driven conditions, which is a surprising and active area of research. In trials where people were openly told their pills were inert and that placebos can help through mind-body processes, groups with conditions like irritable bowel syndrome and chronic back pain reported more improvement than those given nothing. Pooled across those trials the effect is large by the statistical measure, a standardized mean difference around 0.88, but the trials are still few and short, and like concealed placebos they act on symptoms rather than on objective disease.
What is the nocebo effect?
It is the placebo effect in reverse: expecting harm can produce harm. When people who had stopped statins for muscle symptoms were tested against an identical placebo, 90% of their symptoms occurred on the placebo too. Men warned about the sexual side effects of a prostate drug reported them about three times as often as men not warned, 44% versus 15%. The symptoms are felt; the expectation, not only the drug, is generating them.
Why do controlled trials need a placebo group?
Because so many things make a symptom improve regardless of the treatment: the placebo response, the natural easing of an illness, and the tendency to seek help at your worst and then drift back to average. A single person getting better cannot separate those from the treatment. Comparing the treatment against a placebo given in the same setting cancels out expectation and natural history, so whatever advantage remains is the specific effect of the treatment. That is why careful medicine weighs controlled comparisons over testimonials.
Explore Related
Other pages this one connects to, by the evidence they share, the outcomes they touch, and the ground they cover.
How this connects
- Related
All 11 sources on this page independently checked and cross-referenced.
Thomas Dehli, Founder & Editor, Sacred Lotus
Sacred Lotus has published Chinese medicine reference material since 2001. Integrative pages are held to the same standard as the herb and formula library: cite the source, grade the claim at its real strength, and say where the research has not looked. This page is educational and it is not medical advice. Last reviewed and updated August 10, 2026.