How Breathing Research Is Done and Its Limits: Blinding, Controls and Self-Report

Why this page matters

Headlines about breathwork often sound certain. The studies behind them are mostly small, short and hard to blind, and the reviews that pool them say so themselves. The authors of one major review of breathwork and stress urged caution and better-designed trials so that hype does not run ahead of evidence. [1] This page explains how breathing research is done so you can read a claim with realistic expectations. It is general information, not medical advice.

The main kinds of breathing studies

• Lab physiology studies. Healthy volunteers breathe at a set pace while heart rate, blood pressure or brain activity is recorded. A meta-analysis of voluntary slow breathing included 223 studies: 172 measured effects during the breathing session, 16 right after one session and 49 after a multi-session program. [2] These show what the body does, not whether people feel or live better.
• Randomized trials of programs. People are assigned by chance to a breathing program or a comparison. A pooled review of breathwork and self-reported stress found 12 such trials with 785 adults. [1]
• Systematic reviews and meta-analyses. These pool trials, so they are only as good as the trials inside them. That review rated most of its studies as moderate risk of bias. [1] A review of diaphragmatic breathing (48 trials) found that only 2.12% of outcomes were at low risk of bias. [3]

Why blinding is the hardest problem

In a drug trial, a sugar pill looks the same as the real pill. There is no "placebo breath" that people cannot tell from the real practice. A Cochrane review of breathing exercises in COPD found that few studies reported assessor blinding, and it warned that effects on patient-reported outcomes may have been overestimated because of the lack of blinding. [4]

Expectation can matter a great deal. In a trial of people with anxiety attacks, two breathing therapies with opposite instructions (one aimed to raise carbon dioxide, the other to lower it) both helped, and patients improved equally. [5] The authors suggested that non-specific factors played a role, and patient confidence that the therapy would work accounted for almost half of the variance in improvement at six months. [5] A commentary replied that the design and size of that work did not justify concluding that the two protocols are equally effective. [6] Both readings are fair: expectation is powerful, and one small trial cannot settle the question.

What the comparison group is

The result depends on what people are compared with.

• Waitlist or no treatment. People who wait with no help may not improve, which can flatter any program. In depression trials of psychotherapy, effect sizes differed substantially depending on the control condition used. [7] A review in anxiety disorders reported much higher effect sizes against waitlist controls than against psychological placebos, and noted waitlists are used in up to 73% of published psychological treatment studies. [8] Those papers are about psychotherapy. A breathing-specific analysis of this question was not described in the sources we read.
• Active comparison. In the breathwork and stress review, seven studies had inactive and five had active control groups, and effects were significant for both. [1] In pooled data from five experiments with 245 healthy, pain-free volunteers, mindfulness meditation reduced experimentally induced pain more than a slow-breathing "sham" meditation or a book-listening control. [9] The comparison you choose changes the answer.
• Education or usual care. In the Cochrane asthma review, 20 studies used an inactive control and only two used asthma education. [10]

Self-reported versus measured outcomes

Many techniques, few standards

Diaphragmatic breathing trials ranged from 2 to 10 breaths per minute, 3 to 45 minutes per session, and one session to 12 weeks. [3] A review of dysfunctional breathing found intervention components highly variable and inadequately reported, with low-quality evidence for efficacy. [13] In the asthma review, 14 of 22 studies used yoga and others used retraining, Buteyko, Papworth or diaphragmatic breathing; incomplete reporting hampered the risk-of-bias assessment. [10] A 2025 review describes the field as limited by inconsistent study quality and methodological heterogeneity. [14] "Breathwork" results are therefore a mixture of different practices.

Missing studies and missing harms

The breathwork and stress review checked for publication bias. Its funnel plot suggested some skew from small studies, but a statistical test was non-significant, and 69 further null studies would be needed to erase the stress finding. [1] These checks work from published studies, so they cannot show what was never reported.

Harms are often not measured. The Cochrane asthma review did not include adverse effects as an outcome. [10] The diaphragmatic breathing review called safety underreported (18.75% of studies) and noted no serious adverse events. [3] In the stress review, only four studies (six counting secondary-outcome studies) actively reported on adverse events. [1] Its authors called for more reporting and for research on the safety of fast-paced breathwork in particular. [1]

Some breathing methods are taught or sold by the people who study them. We did not check author ties for every study cited here. When you read a study, check its funding and conflict-of-interest statements.

A short checklist for any breathing claim

1. Which exact practice, at what pace, for how long?
2. Compared with what: waitlist, usual care or an active comparison?
3. Was the outcome reported by people or measured?
4. Were participants healthy volunteers or people with the condition?
5. How many people, and how long were they followed?
6. Were harms reported?
7. Who ran and paid for the study?

Evidence labels in practice

1. Does slow breathing raise vagally mediated heart-rate variability in healthy people?

   Established as a short-term physiological finding

   Increases during and after sessions across 223 studies [2]

2. Does breathwork lower self-reported stress?

   Promising but limited

   Small-to-medium effect, moderate risk of bias [1]

3. Does slow breathing lower blood pressure?

   Promising but limited

   Modest reduction, short follow-up [11][12]

4. Does breathwork cure or prevent disease?

   Insufficient evidence

   Proper evidence not found. We checked NCCIH, the Cochrane asthma and COPD reviews and the pooled reviews cited here; they measure symptoms, quality of life and risk markers, not cures. [12]