Placebo

A person with a headache swallows a sugar pill they believe is a painkiller. Half an hour later, the headache has faded. What happened?

There are at least four possible explanations, and only one of them involves the pill. The headache may have been going away anyway. The person may have taken the pill at the worst moment of the pain, so almost any later moment would feel better. They may report feeling better partly to be polite, or because they expect to. Or the act of taking what they believed was medicine may have changed how their nervous system processed pain.

Only the last of these is a placebo effect in the strict sense. The difficulty — and the reason the subject has been argued over for seventy years — is that from the outside, all four look the same.

Separating placebo effects from their look-alikes reveals both a measurable biology and a firm limit: treatment context can change some symptoms, but it does not cure every disease that can make a person feel ill.

Response Versus Effect

The word placebo is Latin for "I shall please". For centuries it described remedies given more to comfort a patient than to cure them. In the twentieth century it gained a technical meaning: an inert treatment, such as a sugar pill, saline injection, or sham procedure, used as a comparison in clinical trials.

Researchers now separate two terms that are often confused.

• Placebo response: everything that happens to people in the placebo group of a trial. It includes natural recovery, statistical artefacts, reporting bias, and any genuine placebo effect.
• Placebo effect: the part of that improvement caused by the treatment context — the pill, the injection, the white coat, the ritual of care, and the expectations these create. ^[1]

The simplest way to see the distinction is to write down what a trial measures.

The Formal Description

Suppose a trial has three groups: people given an active drug, people given an identical-looking placebo, and people given no treatment at all. Each group's average improvement, $\Delta$, can be broken into parts:

where:

• $N$ is the natural history of the condition — many illnesses improve or fluctuate on their own;
• $R$ is regression to the mean — people usually enter trials when symptoms are at their worst, so later measurements tend to be less extreme even without any intervention; ^[2]
• $B$ is bias in measurement or reporting, such as participants' wish to please researchers;
• $P$ is the placebo effect — the change caused by the treatment context and the expectations it creates;
• $S$ is the specific effect of the drug's chemistry.

A standard two-arm trial compares the drug group with the placebo group, so it measures only $\Delta_{\text{drug}} - \Delta_{\text{placebo}} = S$. This is exactly what drug regulators want to know, but it says nothing about the size of $P$. To estimate the placebo effect itself, a trial needs a no-treatment group:

This small piece of algebra explains most of the history of the subject. For decades, the whole placebo response was interpreted as a placebo effect, which made placebos look far more powerful than they are.

In practice, the equation is a simplification. It assumes these components simply add together, and some evidence suggests expectation can interact with a drug's chemistry, making an active drug more or less effective depending on what the patient believes. ^[1]

Historical Development

The Powerful Placebo

In 1955 the Harvard anaesthesiologist Henry Beecher published "The Powerful Placebo" in the Journal of the American Medical Association. Pooling 15 studies with about 1,000 patients, he concluded that roughly 35% of patients were satisfactorily relieved by a placebo alone. ^[3] The paper helped establish the placebo-controlled trial as the standard of modern medicine, and the "one-third respond to placebo" figure became one of the most repeated numbers in medicine.

The figure did not survive re-examination. In 1997 two researchers reanalysed Beecher's sources and found that the improvements he attributed to placebo could be explained by natural recovery, regression to the mean, additional treatments given alongside the placebo, and other artefacts. None of his source studies included a no-treatment group, so none could separate $P$ from $N + R + B$. ^[4] Beecher had measured the placebo response and called it the placebo effect.

Is the Placebo Powerless?

In 2001, Asbjørn Hróbjartsson and Peter Gøtzsche published a systematic review of trials that included both a placebo group and a no-treatment group — exactly the design needed to estimate $P$. ^[5] Across 130 trials, they found no significant placebo effect on binary outcomes (such as "cured / not cured") or on objective measurements. They did find small effects on continuous, subjective outcomes, particularly pain.

Their updated Cochrane review in 2010, covering 202 trials across 60 clinical conditions, reached a similar conclusion. ^[6] Placebo interventions had no clinically important effects in general, but produced small benefits on patient-reported outcomes, especially pain (standardised mean difference about $-0.28$) and nausea (about $-0.25$). Effects were larger with physical placebos such as sham acupuncture, in small trials, and in trials designed specifically to study placebo. The authors noted that it was difficult to distinguish real patient-reported effects from biased reporting.

The balanced reading of this literature is that placebo effects are real but mostly modest, and concentrated in how symptoms are experienced rather than in the underlying disease.

How It Works

If placebo effects on pain are real, something in the nervous system must produce them. Over the past five decades researchers have identified two main routes — expectation and conditioning — and several of the chemical systems they act through.

Expectation

The first route is belief about what will happen. Verbal information, the appearance of a treatment, and the manner of the clinician all shape expectations, and expectations shape symptoms.

Several classic studies show how much the details matter. Women with headaches who took a placebo tablet labelled with a well-known brand name reported more relief than those who took an unbranded placebo. ^[7] A systematic review found that the colour of pills influences their perceived effect: blue and green tablets tend to be experienced as calming, red, yellow, and orange as stimulating. ^[8] More elaborate and invasive placebos, such as sham procedures, tend to produce larger effects than pills. ^[6]

The relationship with the clinician also matters. In a trial of 262 patients with irritable bowel syndrome, adequate symptom relief was reported by 28% of those on a waiting list, 44% of those given sham acupuncture with a brief, neutral interaction, and 62% of those given sham acupuncture with a warm, attentive, and confident practitioner. ^[9] The needles were identical; the care was not. This connects placebo research to broader questions about how trust shapes human interaction.

Recent work shows that expectation can scale like a dose. Nicotine-dependent smokers vaped an e-cigarette containing the same amount of nicotine, but were told it was low, medium, or high strength. Brain responses in the thalamus, a region rich in nicotine receptors, increased with the believed dose, not the actual one. ^[10] Every participant received real nicotine; the belief changed how the brain responded to it.

Conditioning

The second route is learning by association, the process Ivan Pavlov demonstrated when dogs learned to salivate at a signal that had repeatedly preceded food. If a drug has been taken several times in a particular context, the context alone can come to trigger part of the drug's effect.

An elegant 2003 experiment separated the two routes. ^[11] Healthy volunteers were told that an injection would raise or lower two hormones, growth hormone and cortisol. Verbal suggestion alone had no effect on either hormone. Then, over two days, they received sumatriptan, a drug that raises growth hormone and lowers cortisol. When they were later given a saline injection, their hormones moved in the same direction as with the drug — even when they had been told to expect the opposite. For hormones, the body had learned from the pairing of injection and drug, regardless of what participants consciously believed.

Pain and movement behaved differently in the same study. In those systems, verbal suggestions could override prior conditioning. The conclusion was that placebo responses involving conscious functions such as pain are driven mainly by expectation, while those involving unconscious physiological functions such as hormone secretion are driven mainly by conditioning. ^[11]

Endogenous Opioids

The first biochemical clue came in 1978. Patients recovering from dental surgery were given a placebo they believed was a painkiller. Some reported substantial relief. When those same patients were then given naloxone, a drug that blocks opioid receptors, their pain increased. ^[12] The implication was that placebo relief was being produced, at least in part, by the brain's own opioid-like chemicals — endorphins.

Later work refined this. Placebo analgesia produced by expectation alone could be blocked by naloxone, but placebo analgesia produced by conditioning with a non-opioid painkiller was only partly blocked, pointing to more than one chemical pathway. ^[13] Brain imaging showed that placebo treatment increased prefrontal activity while people anticipated pain and reduced activity in pain-processing regions such as the thalamus, insula, and anterior cingulate cortex during pain. ^[14] A later study showed that placebo analgesia engaged a descending pain-control pathway, from the cortex through the brainstem periaqueductal gray, and that naloxone disrupted it. ^[15] This is the same system the body uses to suppress pain during danger or intense exertion.

Dopamine in Parkinson's Disease

Parkinson's disease is caused by the loss of dopamine-producing neurons that are needed for smooth movement. In 2001, researchers used positron emission tomography (PET) with a radioactive tracer, raclopride, that competes with dopamine for the same receptors. When patients with Parkinson's received a placebo they believed might be an active drug, less tracer bound to receptors in the striatum — indicating that the patients' own brains had released substantial amounts of dopamine. ^[16]

This result matters for two reasons. It showed a placebo altering a measurable neurochemical in a disease with clear biological damage, and it suggested that dopamine, the brain's signal of expected reward, may be part of the machinery of expectation in general. It does not mean that placebo can treat Parkinson's disease: the effect is smaller and less reliable than that of dopamine-replacing drugs, and it cannot stop the neurons from degenerating.

Predictive Processing

A unifying framework has emerged from computational neuroscience. The brain is described not as a passive receiver of signals but as a prediction machine. It constantly combines incoming sensory information with prior expectations to produce its best estimate of what is happening. ^[17,18]

In this view, a pain experience is the brain's estimate of the body's state, weighted by how reliable each source of information seems. If you strongly expect relief and the incoming pain signal is ambiguous, the estimate shifts toward the expectation. If the signal is intense and unambiguous — a broken bone, a blocked airway — there is less room for expectation to change the estimate.

This framework is an accepted theoretical model that explains many findings and makes testable predictions; placebo effects should be larger when symptoms are ambiguous and expectations are confident. It is not yet a complete mechanistic account.

Other Proposed Pathways

Popular accounts, including a widely viewed podcast episode on placebo and belief effects, ^[29] sometimes cite research on a pathway from the prefrontal cortex to the hypothalamus that controls stress responses such as heart rate and body temperature. That work, done in rats, identified a genuine cortical route from social stress to the body's physiology. ^[19] It shows how thoughts and social context could reach basic bodily functions. Whether this specific pathway contributes to placebo effects in humans has not been tested; linking the two is speculation.

Similarly, variants of a gene called COMT, which affects how dopamine is broken down, have been associated with the size of placebo responses in irritable bowel syndrome. ^[20] This has launched a research program on the genetics of placebo response, ^[21] but the findings are early, based on small samples, and not yet reliable enough to predict who will respond.

Nocebo: The Dark Twin

If expecting benefit can reduce symptoms, expecting harm can create them. This is the nocebo effect, from the Latin "I shall harm". ^[1]

Nocebo effects follow similar pathways. Anxiety-driven nocebo pain has been linked to cholecystokinin, a chemical messenger that opposes opioid pain relief; blocking it with the drug proglumide prevented nocebo-induced increases in pain after surgery. ^[22]

The clinical consequences can be large. In the SAMSON trial, 60 patients who had stopped taking statins because of side effects each took a year of monthly bottles containing a statin, a placebo, or nothing, in random order. Symptoms were real and often severe, but about 90% of the symptom burden during statin months also appeared during placebo months. ^[23] Half the patients later successfully resumed their statins. Their side effects were not imaginary; they were largely caused by the act of taking tablets and the expectation of harm, not by the drug.

Nocebo effects create an ethical tension. Informed consent requires warning patients about side effects, but the warning itself can produce some of them.

Open-Label Placebo

Most people assume placebos only work if the patient is deceived. In 2010, a team at Harvard tested this assumption directly. Eighty patients with irritable bowel syndrome were randomised either to no treatment or to pills openly described as placebos — "like sugar pills" — along with an explanation that placebo pills had produced improvement in clinical studies. After three weeks, the open-label placebo group reported significantly greater global improvement and symptom relief. ^[24]

A 2021 meta-analysis of 11 trials found that open-label placebos produced medium-to-large improvements on self-reported outcomes in conditions including back pain, cancer-related fatigue, irritable bowel syndrome, and depression. ^[25]

These results are intriguing, but caution is needed. Patients cannot be blinded to receiving an open-label placebo, outcomes were largely subjective, and the explanation participants receive is itself a powerful expectation-building intervention. The trials show that deception is not necessary for placebo responses. They do not show that sugar has healing properties.

What Placebo Can and Cannot Change

The evidence consistently points to a clear boundary.

A striking example comes from asthma. In a 2011 study, patients received an albuterol inhaler, a placebo inhaler, sham acupuncture, or no intervention, in rotation. ^[26] Albuterol improved lung function (measured as the volume of air forcibly exhaled in one second) by about 20%; the placebo inhaler and sham acupuncture by about 7%, the same as no intervention. Yet patients reported improvement at similar rates with albuterol (50%), placebo inhaler (45%), and sham acupuncture (46%), compared with 21% for no intervention.

The placebo changed how breathing felt. It did not change how the lungs worked. For a patient in an asthma attack, that distinction could be dangerous.

The same pattern appears in cancer. A review of placebo-controlled cancer trials found that placebos sometimes improved pain and appetite, but tumour shrinkage meeting standard response criteria occurred in only 10 of 375 patients given placebo — about 3%, a rate that cannot be clearly separated from spontaneous regression or measurement variation. ^[27] There is no credible evidence that expectation can shrink tumours, clear bacterial infections, repair broken bones, or restore dead neurons.

The practical rule that emerges:

• Can change (often modestly): pain, nausea, fatigue, subjective breathlessness, perceived symptom severity, some hormonal and neurochemical responses, and the side effects attributed to drugs. ^[6,11,16,23]
• Cannot change, according to current evidence: tumour size, infection, lung function in asthma, structural damage, or the underlying course of most diseases. ^[26,27]

Placebo effects work by adjusting how the brain predicts, attends to, and interprets the body's signals, and through learned physiological responses. They operate within the reach of the nervous system and its connections to the body, not beyond it.

Placebo Responses Are Rising

A curious recent finding complicates clinical trials. An analysis of randomised trials of drugs for chronic neuropathic pain between 1990 and 2013 found that placebo responses increased substantially over that period while drug responses stayed stable, shrinking the apparent advantage of the drugs. ^[28] The increase was seen only in trials conducted in the United States, where trials also became larger and longer.

Why this happened is unclear. Possible explanations include changes in how participants are recruited, the growing length and intensity of trial contact, and changes in public expectations shaped by direct-to-consumer drug advertising. None has been confirmed. Whatever the cause, it has made it harder for genuinely effective new drugs to show their benefit.

Limitations and Open Problems

• Effect size. Outside of pain, nausea, and a few other subjective outcomes, placebo effects measured against no treatment are usually small. ^[5,6]
• Bias. Subjective outcomes are vulnerable to reporting bias, and it is often impossible to know how much of an apparent placebo effect is genuine change in experience. ^[6]
• Variability. Placebo effects differ widely between individuals and settings, and no reliable method yet predicts who will respond. ^[21]
• Mechanisms. The pain and dopamine pathways are well studied. For most other conditions, the biology remains unclear.
• Ethics. Deceptive placebo use conflicts with informed consent. Open-label placebo is one proposed solution, but its effects outside subjective symptoms are unproven.
• Popular exaggeration. The claim that "the mind can heal anything" is not supported. The more accurate claim is narrower and more interesting: expectation and learning can change how the nervous system processes certain signals.

Summary of Certainty

• Established fact: placebo groups in trials improve, mostly because of natural history, regression to the mean, and bias. ^[2,4]
• Established fact: genuine placebo effects on pain exist and involve endogenous opioid systems. ^[12,13,15]
• Well-supported: expectation can release dopamine in Parkinson's disease; conditioning can reproduce hormonal drug effects. ^[11,16]
• Well-supported: placebo effects are modest overall and concentrated on subjective outcomes. ^[5,6]
• Accepted theoretical model: predictive processing explains placebo effects as expectation-weighted perception. ^[17,18]
• Promising but uncertain: open-label placebos; genetic predictors of placebo response. ^[20,24,25]
• Speculation: links between specific stress circuits in animals and human placebo effects. ^[19]

Why the Difference Matters

The placebo effect sits at the point where the mind meets the body. It shows that beliefs, context, and learning are not abstractions floating above physiology; they act through real chemical and neural pathways. The same research also shows the limits of that power with unusual clarity: expectation can ease pain and change brain chemistry, but it cannot shrink a tumour or open an airway.

For medicine, both lessons matter. Every drug must prove itself against placebo, because the ritual of treatment has effects of its own. Every clinician works with placebo and nocebo effects whether they intend to or not, through the words they use, the confidence they show, and the warnings they give. And patients deserve to know both that their expectations can genuinely shape how they feel, and that feeling better is not always the same as being better.

The mechanisms connect to other questions about how attention and interpretation shape experience; see Meditation and Suffering. What remains unresolved is how far the principle extends: which physiological systems are within reach of expectation, why some people respond much more strongly than others, and whether that power can be used deliberately and honestly in treatment.

References

[1] Colloca, L., & Barsky, A. J. (2020). "Placebo and Nocebo Effects." New England Journal of Medicine, 382(6), 554–561. https://doi.org/10.1056/NEJMra1907805

[2] Barnett, A. G., van der Pols, J. C., & Dobson, A. J. (2005). "Regression to the mean: what it is and how to deal with it." International Journal of Epidemiology, 34(1), 215–220. https://doi.org/10.1093/ije/dyh299

[3] Beecher, H. K. (1955). "The Powerful Placebo." Journal of the American Medical Association, 159(17), 1602–1606. https://doi.org/10.1001/jama.1955.02960340022006

[4] Kienle, G. S., & Kiene, H. (1997). "The Powerful Placebo Effect: Fact or Fiction?" Journal of Clinical Epidemiology, 50(12), 1311–1318. https://doi.org/10.1016/S0895-4356(97)00203-5

[5] Hróbjartsson, A., & Gøtzsche, P. C. (2001). "Is the Placebo Powerless? An Analysis of Clinical Trials Comparing Placebo with No Treatment." New England Journal of Medicine, 344(21), 1594–1602. https://doi.org/10.1056/NEJM200105243442106

[6] Hróbjartsson, A., & Gøtzsche, P. C. (2010). "Placebo interventions for all clinical conditions." Cochrane Database of Systematic Reviews, CD003974. https://doi.org/10.1002/14651858.CD003974.pub3

[7] Branthwaite, A., & Cooper, P. (1981). "Analgesic effects of branding in treatment of headaches." BMJ, 282(6276), 1576–1578. https://doi.org/10.1136/bmj.282.6276.1576

[8] de Craen, A. J. M., Roos, P. J., de Vries, A. L., & Kleijnen, J. (1996). "Effect of colour of drugs: systematic review of perceived effect of drugs and of their effectiveness." BMJ, 313(7072), 1624–1626. https://doi.org/10.1136/bmj.313.7072.1624

[9] Kaptchuk, T. J., Kelley, J. M., Conboy, L. A., et al. (2008). "Components of placebo effect: randomised controlled trial in patients with irritable bowel syndrome." BMJ, 336, 999–1003. https://doi.org/10.1136/bmj.39524.439618.25

[10] Perl, O., Shuster, A., Heflin, M., et al. (2024). "Nicotine-related beliefs induce dose-dependent responses in the human brain." Nature Mental Health, 2(2), 177–188. https://doi.org/10.1038/s44220-023-00188-9

[11] Benedetti, F., Pollo, A., Lopiano, L., Lanotte, M., Vighetti, S., & Rainero, I. (2003). "Conscious Expectation and Unconscious Conditioning in Analgesic, Motor, and Hormonal Placebo/Nocebo Responses." Journal of Neuroscience, 23(10), 4315–4323. https://doi.org/10.1523/JNEUROSCI.23-10-04315.2003

[12] Levine, J. D., Gordon, N. C., & Fields, H. L. (1978). "The mechanism of placebo analgesia." The Lancet, 312(8091), 654–657. https://doi.org/10.1016/S0140-6736(78)92762-9

[13] Amanzio, M., & Benedetti, F. (1999). "Neuropharmacological Dissection of Placebo Analgesia: Expectation-Activated Opioid Systems versus Conditioning-Activated Specific Subsystems." Journal of Neuroscience, 19(1), 484–494. https://doi.org/10.1523/JNEUROSCI.19-01-00484.1999

[14] Wager, T. D., Rilling, J. K., Smith, E. E., et al. (2004). "Placebo-Induced Changes in fMRI in the Anticipation and Experience of Pain." Science, 303(5661), 1162–1167. https://doi.org/10.1126/science.1093065

[15] Eippert, F., Bingel, U., Schoell, E. D., et al. (2009). "Activation of the Opioidergic Descending Pain Control System Underlies Placebo Analgesia." Neuron, 63(4), 533–543. https://doi.org/10.1016/j.neuron.2009.07.014

[16] de la Fuente-Fernández, R., Ruth, T. J., Sossi, V., Schulzer, M., Calne, D. B., & Stoessl, A. J. (2001). "Expectation and Dopamine Release: Mechanism of the Placebo Effect in Parkinson's Disease." Science, 293(5532), 1164–1166. https://doi.org/10.1126/science.1060937

[17] Büchel, C., Geuter, S., Sprenger, C., & Eippert, F. (2014). "Placebo Analgesia: A Predictive Coding Perspective." Neuron, 81(6), 1223–1239. https://doi.org/10.1016/j.neuron.2014.02.042

[18] Ongaro, G., & Kaptchuk, T. J. (2019). "Symptom perception, placebo effects, and the Bayesian brain." Pain, 160(1), 1–4. https://doi.org/10.1097/j.pain.0000000000001367

[19] Kataoka, N., Shima, Y., Nakajima, K., & Nakamura, K. (2020). "A central master driver of psychosocial stress responses in the rat." Science, 367(6482), 1105–1112. https://doi.org/10.1126/science.aaz4639

[20] Hall, K. T., Lembo, A. J., Kirsch, I., et al. (2012). "Catechol-O-Methyltransferase val158met Polymorphism Predicts Placebo Effect in Irritable Bowel Syndrome." PLOS ONE, 7(10), e48135. https://doi.org/10.1371/journal.pone.0048135

[21] Hall, K. T., Loscalzo, J., & Kaptchuk, T. J. (2015). "Genetics and the placebo effect: the placebome." Trends in Molecular Medicine, 21(5), 285–294. https://doi.org/10.1016/j.molmed.2015.02.009

[22] Benedetti, F., Amanzio, M., Casadio, C., Oliaro, A., & Maggi, G. (1997). "Blockade of nocebo hyperalgesia by the cholecystokinin antagonist proglumide." Pain, 71(2), 135–140. https://doi.org/10.1016/S0304-3959(97)03346-0

[23] Wood, F. A., Howard, J. P., Finegold, J. A., et al. (2020). "N-of-1 Trial of a Statin, Placebo, or No Treatment to Assess Side Effects." New England Journal of Medicine, 383(22), 2182–2184. https://doi.org/10.1056/NEJMc2031173

[24] Kaptchuk, T. J., Friedlander, E., Kelley, J. M., et al. (2010). "Placebos without Deception: A Randomized Controlled Trial in Irritable Bowel Syndrome." PLOS ONE, 5(12), e15591. https://doi.org/10.1371/journal.pone.0015591

[25] von Wernsdorff, M., Loef, M., Tuschen-Caffier, B., & Schmidt, S. (2021). "Effects of open-label placebos in clinical trials: a systematic review and meta-analysis." Scientific Reports, 11, 3855. https://doi.org/10.1038/s41598-021-83148-6

[26] Wechsler, M. E., Kelley, J. M., Boyd, I. O. E., et al. (2011). "Active Albuterol or Placebo, Sham Acupuncture, or No Intervention in Asthma." New England Journal of Medicine, 365(2), 119–126. https://doi.org/10.1056/NEJMoa1103319

[27] Chvetzoff, G., & Tannock, I. F. (2003). "Placebo Effects in Oncology." Journal of the National Cancer Institute, 95(1), 19–29. https://doi.org/10.1093/jnci/95.1.19

[28] Tuttle, A. H., Tohyama, S., Ramsay, T., et al. (2015). "Increasing placebo responses over time in U.S. clinical trials of neuropathic pain." Pain, 156(12), 2616–2626. https://doi.org/10.1097/j.pain.0000000000000333

[29] Huberman Lab (n.d.). "How Placebo Effects Work to Change Our Biology & Psychology." YouTube (further watching). https://www.youtube.com/watch?v=gdUNjPijwA8