Research · Explainer

The placebo effect and how expectations can change how we feel

Expectations can alter how a symptom is experienced even when the treatment-specific ingredient has not changed. That influence may come from prior learning, attention, reassurance, warning signals or the ritual of care, and it is most clearly demonstrated for experiences such as pain and nausea rather than as a general cure for disease.

This article distinguishes context-driven effects from all of the other reasons that people improve while assigned to an inactive intervention. It then uses randomised trials, experimental pain studies, neuroimaging and trial-methodology guidance to show where subjective relief, physiological measurements and treatment-specific effects do - and do not - line up.

Patient speaking with a clinician in a consultation room about symptoms and expectations
A patient discusses symptoms and expectations with a clinician during consultation Editorial illustration

Separating contextual change from the total response in a placebo arm

A placebo is useful because it imitates selected features of treatment without containing the therapeutic component whose efficacy is being tested. For a medicine trial, that may mean a look-alike tablet; for a procedure study, it may mean a credible sham. The comparison is designed to keep much of the surrounding study experience similar while changing the component of interest.

Improvement recorded in the inactive-comparator group has several possible causes, and expectation is only one of them. Symptoms can fluctuate naturally, enrolment may coincide with an unusually bad period, repeated measurement can alter reporting, and regression towards a participant's usual level can occur without any contextual mechanism at all. For that reason, the full change from baseline in this group is better described as a response observed under placebo conditions, not as a direct measurement of expectation.

A causal estimate requires a second benchmark that shows what would have happened without the sham procedure. Hróbjartsson and Gøtzsche (2010) restricted their Cochrane analysis to randomised studies containing both an inactive intervention and a no-treatment comparison, drawing outcome data from 202 trials covering 60 clinical conditions. Across the evidence as a whole they did not identify an important clinical benefit attributable to placebo interventions, although some continuous outcomes reported by patients, particularly pain and nausea, showed small and inconsistent differences.

How information and previous treatment experience can reshape symptoms

Meaning and attention can alter the appraisal of bodily signals

What a person expects is not merely an opinion stated before treatment; it can alter which sensations receive attention and how those sensations are interpreted. A reassuring explanation can lower the perceived threat of an incoming signal, whereas a warning can make the same signal more salient. These shifts are most relevant to outcomes that depend partly on perception, such as pain intensity, nausea or breathlessness, because the nervous system must interpret the signal before it becomes a reported experience.

Neuroimaging studies offer evidence that this process is accompanied by measurable changes in brain activity. In the experimental pain study by Wager et al. (2004), volunteers underwent functional MRI while painful stimulation was paired with a sham analgesic procedure. When relief was expected, activity during pain fell in regions including the thalamus, insula and anterior cingulate cortex, while anticipatory activity rose in prefrontal areas. The stimulus had not vanished; the pattern of neural processing differed under an expectation of relief.

The same contextual machinery can amplify discomfort rather than reduce it. Negative information, memories of unpleasant treatment or other threat cues may increase symptom reporting or make an unchanged therapy seem less effective; this family of effects is usually labelled nocebo. Colloca and Barsky (2020) emphasise that placebo and nocebo phenomena vary with circumstances and learning history, so an effect observed in one experiment is not a stable property that can be assumed to recur unchanged elsewhere.

Learned associations can recruit different physiological pathways

Past treatment can leave associations that influence a later response even when the later procedure is pharmacologically inactive. Amanzio and Benedetti (1999) compared verbal expectancy with conditioning created by prior exposure to analgesic drugs in an experimental pain model. Naloxone blocked some opioid-linked placebo analgesia, whereas a component conditioned with a non-opioid drug was not blocked in the same way. The result matters because apparently similar behavioural relief can arise through more than one biological route.

Expectancy can also alter the experienced benefit of an active drug rather than act only in the absence of one. Bingel et al. (2011) kept the remifentanil concentration constant during heat-pain testing but changed the information volunteers received about whether analgesia should occur. Positive expectancy roughly doubled the reported analgesic benefit, while negative expectancy abolished it; functional imaging changed alongside the reports. The experiment therefore shows interaction between pharmacology and context, not replacement of pharmacology by belief.

Experimental pain models are especially valuable for identifying mechanism because stimulus intensity, timing, drug exposure and instructions can be tightly controlled. That control strengthens causal interpretation, but it also limits generalisation to routine care, where illness, clinician behaviour, treatment history and follow-up are more complicated. These experiments establish that contextual information can modify pain processing; they do not provide a universal percentage that can be applied to every symptom or clinical setting.

Clinical studies show that context matters differently across outcomes

The clinical encounter can change what patients report

Kaptchuk et al. (2008) tested the contribution of the treatment encounter in 262 adults with irritable bowel syndrome by assigning them to a waiting list, sham acupuncture with limited interaction, or sham acupuncture delivered with an augmented clinician interaction. After three weeks, adequate relief was reported by 28%, 44% and 62% of participants in those groups respectively. Because the inactive procedure was similar in the last two groups while the interaction differed, the gradient is evidence that the encounter itself can contribute to symptom reports.

Those percentages should not be treated as a general estimate of placebo responsiveness. The trial examined one condition over a short interval and relied heavily on patient-reported relief; it did not show that the same increase would occur in another disorder, with another ritual or with an objective endpoint. Its strength is narrower and more useful: it demonstrates experimentally that the way care is delivered can become part of the measured outcome.

Even the form of the inactive intervention can influence study results. Hróbjartsson and Gøtzsche (2010) found substantial variation between trials, including within pain studies judged at lower risk of bias, and their review suggested larger average effects for some physical sham procedures and for outcomes based on patients' own assessments. That pattern is compatible with a role for treatment ritual, but it may also reflect differences in design, blinding and bias. A sham injection, a tablet and an extended consultation therefore should not be assumed to generate interchangeable contextual effects.

Open-label trials test whether deception is necessary

A separate line of research removes concealment altogether by telling participants that they are receiving a placebo. These open-label studies ask whether symptom change can occur when the inert status of the intervention is explicit, rather than when participants believe they may be receiving an active treatment. They are useful for testing the role of ritual, explanation and learned associations under conditions that do not depend on deception.

Fendel et al. (2025) synthesised 60 randomised controlled trials involving 4,648 participants overall; the main pooled analysis comprised 63 comparisons and included 4,554 participants. The pooled effect across health-related outcomes was small, with a standardised mean difference of 0.35 in favour of open-label placebo conditions. The analysis also separated outcome types: self-reported measures showed an average SMD of 0.39, whereas objectively measured outcomes produced an SMD of 0.09. That contrast argues against treating all apparent placebo benefit as one uniform phenomenon.

The same meta-analysis reported moderate heterogeneity, with I2 = 53% for the overall estimate (Fendel et al., 2025). Effects varied across populations, outcomes and study arrangements, and the included trials differed in quality and risk of bias. Open-label placebo research therefore supports the possibility of modest benefits in some settings, but it does not establish that openly administered inert treatments reliably improve every condition or every type of endpoint.

Why symptom relief and physiological change can point in different directions

A rating of how someone feels and a laboratory or physiological measurement answer different questions. Perceived breathlessness, pain or fatigue can improve even when a disease-specific marker changes little; conversely, a biological measure may shift before the person notices much difference. Treating one type of endpoint as a proxy for the other can therefore obscure what a trial actually measured.

Study

Design and participants

Outcome examined

Finding used in this article

Kaptchuk et al. 2008

262 adults with IBS assigned to three study conditions

Adequate relief at three weeks

Relief was reported by 28% on the waiting list, 44% after a limited encounter and 62% after an augmented encounter

Wechsler et al. 2011

46 randomised participants; 39 completed repeated asthma sessions

FEV1 and perceived improvement

FEV1 increased by roughly 20% after albuterol and about 7% in the other conditions; perceived improvement was 50%, 45%, 46% and 21%

Fendel et al. 2025

60 RCTs, 4,648 participants overall; main pooled analysis: 63 comparisons, n = 4,554

Health outcomes after openly disclosed placebo use

Overall SMD 0.35; SMD 0.39 for self-reports and 0.09 for objective measures

The asthma experiment by Wechsler et al. (2011) illustrates the distinction. Forty-six participants were randomised and 39 completed repeated sessions involving active albuterol, a placebo inhaler, sham acupuncture and no intervention. FEV1 rose by about 20% after albuterol but by only about 7% in each of the other conditions. Patient-rated improvement, however, was 50% after albuterol, 45% after the placebo inhaler, 46% after sham acupuncture and 21% with no intervention.

These findings do not make the reported relief irrelevant; they show that subjective experience and airway physiology are separate outcomes. A person can genuinely feel less breathless without showing the bronchodilation produced by albuterol. The same distinction appears in the 2025 open-label meta-analysis, where pooled effects were clearly larger for self-reported outcomes than for objective measurements. Placebo-related change can therefore be real in experience without demonstrating equivalent change in the disease process itself.

Context can alter symptoms and measurable responses, but those changes are outcome-specific. Relief reported by a patient is not automatically evidence that the underlying pathology has improved.

How placebo controls help isolate treatment-specific effects

Matched study conditions reduce differences beyond the treatment itself

In a placebo-controlled trial, the active-treatment and comparator groups can be exposed to the same schedule of visits, monitoring, explanations and uncertainty about assignment. The intended difference is the treatment-specific component under investigation. Comparing outcomes between those groups estimates the added effect of that component against a background in which many features of participation are shared.

That contrast is not a clean measurement of a single, pure placebo mechanism. Without a no-treatment group, changes observed in the comparator arm still contain natural symptom variation, regression towards the mean, measurement effects and contextual influences. The 2010 Cochrane review used no-treatment comparisons precisely to separate some of those time-related changes from effects associated with receiving a placebo intervention.

For this reason, calling every improvement in an inactive-comparator group a placebo effect overstates what a conventional two-arm trial can show. Its main question is usually whether the active intervention outperforms the matched comparator under otherwise similar trial conditions. A design that also includes a no-treatment arm can address an additional question: whether exposure to the placebo procedure itself changes the outcome compared with receiving no such procedure.

Randomisation and comparator choice protect different parts of the inference

Random allocation deals with who enters each group rather than with the placebo mechanism itself. Altman and Bland (1999) explained that randomisation prevents prognosis, clinician preference or participant characteristics from systematically steering assignment, thereby making the groups comparable at baseline on average. Blinding and comparator selection then address different sources of bias during and after treatment. These design features work together, but they are not interchangeable.

Regulatory guidance frames comparator choice as a question about the conclusion a trial needs to support. ICH E10, issued in 2001 and published by the US Food and Drug Administration, describes how placebo, active-control and other designs answer different evidentiary questions. A placebo comparator can be methodologically powerful because it helps preserve blinding and estimate treatment-specific benefit, but its usefulness depends on the clinical context.

No universal rule makes an inactive comparator appropriate in every study. ICH E10 links the choice to factors such as whether effective therapy already exists, how serious the condition is, which endpoint is being measured and what harm could follow from delaying established treatment. The scientifically cleanest contrast is not automatically the ethically acceptable one; a valid design must answer the research question without exposing participants to an unreasonable comparison.

What the evidence supports and what remains uncertain

Several conclusions are supported across experimental approaches. Wager et al. (2004), Amanzio and Benedetti (1999) and Bingel et al. (2011) show that expectation and learned associations can modify pain-related processing and can interact with active analgesic treatment. Kaptchuk et al. (2008) additionally demonstrates that features of a clinical encounter can alter patient-reported outcomes, while Kaptchuk and Miller (2015) frame placebo effects as responses shaped by the therapeutic context rather than by an inert intervention alone. Together, these findings make contextual influences measurable phenomena rather than a shorthand for imagined symptoms.

The same literature also sets clear boundaries. Hróbjartsson and Gøtzsche (2010) found no important overall clinical effect across the many conditions they reviewed, and results varied markedly by endpoint and study. Wechsler et al. (2011) and Fendel et al. (2025) likewise show that changes in reported experience can exceed changes in objective physiology. Feeling better can therefore be clinically meaningful while still leaving a disease-specific biological process largely unchanged.

The most defensible conclusion is conditional rather than sweeping. Expectations, prior learning and treatment context can alter how some symptoms are experienced and can influence measurable neural or drug-response processes, but neither the size nor the pathway of those effects is fixed across conditions. In clinical research, the main value of a placebo comparator is methodological: it helps separate the effect specific to the intervention being tested from the many other influences that accompany receiving care.

Frequently asked questions

Can a placebo effect produce physical changes in the body?

Yes, some experiments show measurable physiological changes associated with placebo-related expectations, including altered activity in pain-processing brain regions and endogenous opioid signalling. That does not mean every organ or disease process responds to placebo. Objective effects vary by mechanism, condition and outcome, and are often smaller than changes in self-reported symptoms.

Does the placebo effect work when people know they are taking a placebo?

Open-label placebo trials suggest that some benefits can occur even when participants are explicitly told the intervention is a placebo. A 2025 meta-analysis found small average benefits across 60 randomised trials, with larger effects on self-reported than objective outcomes. The findings are heterogeneous and do not establish effectiveness for every condition.

Is feeling better on a placebo the same as the disease getting better?

No. Symptom experience and disease-specific physiology are different outcomes. In a 2011 asthma study, placebo interventions produced substantial self-reported improvement but did not improve lung function beyond the change seen with no intervention, whereas active albuterol clearly increased FEV1. Both types of measurement can matter, but they answer different questions.

Why do placebo effects vary so much between studies?

Placebo effects depend on the outcome being measured, the condition, prior learning, expectations, the treatment ritual, patient-clinician interaction and study design. Trials also differ in blinding, sample size and control groups. The 2010 Cochrane review found marked variation even among pain studies, so there is no single universal placebo-effect size.

Can negative expectations make symptoms feel worse?

Yes. Negative expectations can contribute to nocebo effects, in which symptoms feel worse or a treatment seems less effective because of context and expectation. In a 2011 remifentanil experiment, negative expectancy abolished the drug's reported analgesic benefit under controlled conditions. Such findings show modulation of experience, not a universal response in all illnesses.

Sources

  1. Hróbjartsson A, Gøtzsche PC. Placebo interventions for all clinical conditions. Cochrane Database Syst Rev. 2010;2010(1):CD003974. DOI: 10.1002/14651858.CD003974.pub3.
  2. Kaptchuk TJ, Miller FG. Placebo Effects in Medicine. N Engl J Med. 2015;373(1):8–9. DOI: 10.1056/NEJMp1504023.
  3. Colloca L, Barsky AJ. Placebo and Nocebo Effects. N Engl J Med. 2020;382(6):554–561. DOI: 10.1056/NEJMra1907805.
  4. Wager TD, Rilling JK, Smith EE, et al. Placebo-induced changes in FMRI in the anticipation and experience of pain. Science. 2004;303(5661):1162–1167. DOI: 10.1126/science.1093065.
  5. Amanzio M, Benedetti F. Neuropharmacological dissection of placebo analgesia: expectation-activated opioid systems versus conditioning-activated specific subsystems. J Neurosci. 1999;19(1):484–494. DOI: 10.1523/JNEUROSCI.19-01-00484.1999.
  6. Bingel U, Wanigasekera V, Wiech K, et al. The effect of treatment expectation on drug efficacy: imaging the analgesic benefit of the opioid remifentanil. Sci Transl Med. 2011;3(70):70ra14. DOI: 10.1126/scitranslmed.3001244.
  7. Kaptchuk TJ, Kelley JM, Conboy LA, et al. Components of placebo effect: randomised controlled trial in patients with irritable bowel syndrome. BMJ. 2008;336(7651):999–1003. DOI: 10.1136/bmj.39524.439618.25.
  8. Wechsler ME, Kelley JM, Boyd IOE, et al. Active albuterol or placebo, sham acupuncture, or no intervention in asthma. N Engl J Med. 2011;365(2):119–126. DOI: 10.1056/NEJMoa1103319.
  9. Fendel JC, Tiersch C, Sölder P, Gaab J, Schmidt S. Effects of open-label placebos across populations and outcomes: an updated systematic review and meta-analysis of randomized controlled trials. Sci Rep. 2025;15(1):29940. DOI: 10.1038/s41598-025-14895-z.
  10. Altman DG, Bland JM. Statistics notes. Treatment allocation in controlled trials: why randomise? BMJ. 1999;318(7192):1209. DOI: 10.1136/bmj.318.7192.1209.
  11. US Food and Drug Administration. E10 Choice of Control Group and Related Issues in Clinical Trials. May 2001. https://www.fda.gov/regulatory-information/search-fda-guidance-documents/e10-choice-control-group-and-related-issues-clinical-trials
Adrian Foster

Adrian Foster — author

Adrian Foster writes with an editorial focus on experimental science, laboratory research, controlled comparisons, measurement techniques and the role of replication in evaluating scientific findings. At Cabdirect, Adrian covers experiments, laboratory methods...

Related