I spent two days this month listening to meditation researchers talk to each other. Not to journalists, not to students, not to people deciding whether to take up a practice. To each other.

The register is completely different from the one you meet in an article about meditation, and I think the difference is worth writing down. In public the field speaks in findings. Among themselves they spend a surprising amount of time on what did not work, what did not replicate, and what their instruments cannot see. Several of the talks I found most useful led with a null.

That is a healthy field doing its job. It is also, if you only ever read the press coverage, completely invisible to you.

The published record is 87% positive

Start with the number that frames everything else.

Coronado-Montoya and colleagues went through 124 published randomised controlled trials of mindfulness-based mental health interventions and asked a simple question: how many reported a positive result? The answer, published in PLoS ONE in 2016, was 108 of them — 87%. By the trials’ own stated conclusions, 109 of 124 concluded the intervention worked.

Then they worked out how many should have been positive. Given the effect sizes typically claimed and the statistical power the trials actually had, the expected number was about 65.7, or 53%.

Coronado-Montoya and colleagues, PLoS ONE 2016. The gap between the two bars is the shape of what does not get published.

The observed-to-expected ratio is about 1.6, and it was worse in the smaller, weaker trials. The authors are careful about what this does and does not show: it is consistent with publication bias, with selective outcome reporting, or with both. It is not proof that any particular trial is wrong.

But it does tell you something about what you are reading. If roughly a third of the positive results in the literature should not be there, then the average article summarising “what the research says” is summarising a filtered sample. The nulls were run. Many of them were never printed.

So when a researcher stands up in front of their peers and opens with a failure, they are doing the thing the published record systematically fails to do.

The brain-change story, in two studies

The claim that meditation changes the physical structure of your brain traces largely to one paper, and it is worth looking at it closely, because it is a good paper that has been asked to carry far more than it can.

Hölzel, Carmody, Vangel, Congleton, Yerramsetti, Gard and Lazar published it in Psychiatry Research: Neuroimaging in 2011. Meditation-naive adults did an eight-week mindfulness-based stress reduction course and were scanned before and after; a waitlist group was scanned on the same schedule. Grey matter concentration increased in the left hippocampus, the posterior cingulate cortex, the left temporo-parietal junction, and parts of the cerebellum.

It is a careful study with a real control group. It is also small: sixteen people in the meditation arm after two dropped out, and seventeen controls.

In 2022, Kral, Davis, Korponay, Hirshberg and colleagues, with Richard Davidson, published the bigger version in Science Advances. Two combined randomised controlled trials, 218 meditation-naive adults, randomised three ways: eight-week MBSR, an active non-mindfulness wellness programme matched for time and attention, or no training at all. Grey matter volume, grey matter density and cortical thickness, across the whole brain and in regions of interest taken from the earlier literature.

They found nothing. No structural difference between MBSR and either control group. Their own phrase is that they found no evidence MBSR produced neuroplastic changes, and that they failed to replicate the prior findings.

Bars are drawn to scale, two pixels per participant. The larger, better-controlled study is the one that found nothing — which is the direction these things usually go.

I want to be fair to both papers, because the honest reading is not “the first one was wrong.”

The 2022 authors offer their own explanations. The earlier participants were self-selected into a stress-reduction course, which may mean they had more room to improve; the simple act of choosing to enrol may carry some of the benefit; longer or more meditation-specific training might produce something detectable where eight weeks does not. Any of those could be true.

What the pair of studies does establish is the size of the claim that survives. “Eight weeks of mindfulness reliably thickens your brain” is not supported. It was never supported by a single study of sixteen people, and the attempt to confirm it at scale did not confirm it.

One more detail from the 2022 trial that I keep thinking about. Both the mindfulness group and the active wellness group reported being more mindful than the untrained controls. The authors read that as a sign the self-reported gains may belong to wellness programmes in general rather than to meditation in particular. That is a quieter result than the brain scans and probably a more important one.

The largest school trial found nothing

If you want a single study that shows how the field has matured, it is MYRIAD.

Kuyken and a long list of colleagues ran a cluster randomised trial across 84 schools and 8,376 pupils aged 11 to 14, published in Evidence-Based Mental Health in 2022. Schools were randomised either to a standardised mindfulness curriculum or to teaching as usual. Three co-primary outcomes at one year: risk of depression, social-emotional-behavioural functioning, and well-being.

The standardised mean differences were 0.005, 0.02 and 0.02. All three confidence intervals cross zero. In plain terms: nothing.

Two things stop this being a simple story of failure. The trial did find improvements in school climate and in teacher burnout, though most of those faded within a year. And 80% of pupils did not do the practice they were set, which is a finding about implementation rather than about mindfulness — though it is also, arguably, the most realistic estimate anyone has of what a school mindfulness programme actually delivers when you run it in real schools at scale.

This is the trial that gets cited in the room and almost never in the wellness coverage.

A null the researcher led with

The talk I liked best was built around a result that did not come out.

Lewis-Healey, Kringelbach, Canales-Johnson and Laukkonen have a preprint on whether “critical” brain dynamics — a particular balance between rigid and chaotic activity, measured through long-range temporal correlations in EEG — mark the meditative state. They recorded 33 experienced Transcendental Meditation practitioners, with a mean of nearly thirteen years of practice, and 33 matched controls. Ten minutes of resting EEG, then thirty minutes of meditation for the practitioners and thirty minutes of silently counting upwards for the controls, then ten minutes of resting EEG again.

Compared with the meditators’ own resting state, meditation raised alpha and beta long-range correlations. So far so good.

Then the comparison that matters: meditators meditating versus controls silently counting. No significant difference in any frequency band. Thirteen years of practice, and at the level of this measure the brain was doing what a stranger counting in their head was doing.

What rescued the study was not a better brain measure. It was a better measure of experience. Instead of asking people to rate the session 0 to 10 afterwards, they had participants trace how much effort they had felt, moment by moment, across the thirty minutes. When the brain data were reorganised around that — around felt effortlessness rather than around the label “meditation” — a relationship appeared, and appeared more strongly in the meditators than the controls.

I find this the most instructive result of the lot, and not because of what it says about criticality. It is a demonstration that a null can be a measurement problem on the subjective side rather than the objective one. The state the practitioners were in was real. The category “is meditating” was simply too coarse to find it.

It is also a preprint, which means it has not yet been through peer review.

Things presented but not yet published

Two more results are worth recording, with a clear flag: these were presented at a research meeting, not published. I cannot link you to a paper, you cannot check my summary against one, and both may change before they appear in print. Treat them accordingly.

The first is directly relevant to what I have been writing about here. A group reported a pilot of 90 people, split into three: slow breathing through the left nostril only, through the right nostril only, and normal slow breathing as a control, six seconds in and six seconds out for ten minutes. In the left-nostril group, heart rate variability fell by about 27%, with sympathetic indices rising. The right-nostril group showed no heart rate variability effect, but did show a significant fall in both systolic and diastolic blood pressure. The control group did nothing in particular.

If that holds, it is backwards from the tradition. The left nostril is chandra, the lunar channel, the cooling and quietening one — the one prescribed for gentle work. A pilot that finds it raising sympathetic markers is not a small discrepancy. It is also one unpublished pilot, it runs against other published work on single-nostril breathing, and the presenter called it a proof of concept and asked for more trials. I am recording it because I would want to know, not because it settles anything.

The second is a researcher reporting that her own data on Tibetan monastics came out, in her words, confusing: the monks scored worse on a standard emotion-regulation questionnaire than ordinary people, with more negative and fewer positive emotions. Her reading is that the instrument is the problem rather than the monks — that a questionnaire built on Western assumptions about what good emotion regulation looks like may be measuring the wrong construct in a tradition that emphasises preventing an emotion from arising rather than managing it once it has. She is now building and validating a new instrument from the monks’ own definitions.

I have no way to check either of these. I am telling you that explicitly rather than smoothing it over.

What I take from this

None of this says meditation does not work. The best meta-analysis of the field, by Goyal and colleagues in JAMA Internal Medicine in 2014, pooled 47 trials and 3,515 participants and found moderate evidence of improvement in anxiety, depression and pain — effect sizes around 0.3 to 0.4, which is modest but real. It also found low or insufficient evidence for positive mood, attention, sleep, eating habits, substance use and weight, and no evidence that meditation beat active treatments like drugs or exercise.

That is a smaller claim than the one in circulation, and it has been the accurate summary for over a decade.

The thing I did not expect to come away with is a kind of respect. Van Dam, van Vugt, Vago and a long list of co-authors wrote a paper in Perspectives on Psychological Science in 2018 called “Mind the Hype,” arguing that the practice had outrun the evidence and that poor method and overclaiming would eventually hurt the people the field was trying to help. What struck me this month is how thoroughly that paper seems to have been absorbed. The caution is now the house style, at least when the researchers are among themselves.

The gap is not between the researchers and the truth. It is between the room and the article you read about the room.

So the practical suggestion is boring and I will make it anyway. When you meet a confident claim about what meditation does to your brain, ask how many people were in the study, whether there was a control group that was also doing something, and whether anyone has repeated it with more people. Those three questions would have caught almost everything I have had to correct on this blog.

And sit anyway. The evidence for modest improvements in anxiety and low mood is real, and none of this requires the brain scans to be true.

Sources

1. Coronado-Montoya S, Levis AW, Kwakkenbos L, Steele RJ, Turner EH, Thombs BD. Reporting of positive results in randomized controlled trials of mindfulness-based mental health interventions. PLoS ONE (2016), 11(4):e0153220. doi.org
2. Hölzel BK, Carmody J, Vangel M, Congleton C, Yerramsetti SM, Gard T, Lazar SW. Mindfulness practice leads to increases in regional brain gray matter density. Psychiatry Research: Neuroimaging (2011), 191:36–43. doi.org
3. Kral TRA, Davis K, Korponay C, Hirshberg MJ, Hoel R, Tello LY, Goldman RI, Rosenkranz MA, Lutz A, Davidson RJ. Absence of structural brain changes from mindfulness-based stress reduction: two combined randomized controlled trials. Science Advances (2022), 8(20):eabk3316. doi.org
4. Kuyken W, Ball S, Crane C, et al. Effectiveness and cost-effectiveness of universal school-based mindfulness training compared with normal school provision in reducing risk of mental health problems and promoting well-being in adolescence: the MYRIAD cluster randomised controlled trial. Evidence-Based Mental Health (2022), 25:99–109. doi.org
5. Lewis-Healey E, Kringelbach M, Canales-Johnson A, Laukkonen R. Near-critical brain dynamics track effortlessness during meditation. bioRxiv preprint (2026). biorxiv.org
6. Goyal M, Singh S, Sibinga EMS, Gould NF, Rowland-Seymour A, Sharma R, Berger Z, Sleicher D, Maron DD, Shihab HM, Ranasinghe PD, Linn S, Saha S, Bass EB, Haythornthwaite JA. Meditation programs for psychological stress and well-being: a systematic review and meta-analysis. JAMA Internal Medicine (2014). doi.org
7. Van Dam NT, van Vugt MK, Vago DR, Schmalzl L, Saron CD, Olendzki A, Meissner T, et al. Mind the hype: a critical evaluation and prescriptive agenda for research on mindfulness and meditation. Perspectives on Psychological Science (2018), 13(1). doi.org