How often does meditation go wrong?

Meditation is usually sold as the one health habit with no downside. The research says otherwise, though not in the way the headlines put it.

The numbers

In 2025, Nicholas Van Dam and colleagues (Clinical Psychological Science) surveyed 886 American meditators, matched to the US population on age, gender and ethnicity, and spread across mindfulness, mantra and spiritual practice. Each person worked through a checklist of 30 experiences.

• 58% marked at least one experience as at least mildly unpleasant.
• 31% said yes to a single plain question about a challenging, difficult or distressing experience.
• 9% said a bad effect affected how they functioned day to day.
• 4% said that lasted more than a day.

The press coverage led with the 60% and the 9%. The 4% got less attention.

A separate survey by Goldberg, Lam, Britton and Davidson (Psychotherapy Research, 2021) of 434 meditators found the same pattern: about one in three said yes to the general question, about one in ten reported an effect on functioning, and 6% said it lasted more than a day. The most common effects were anxiety, traumatic memories coming back, and feeling emotionally raw.

Two samples with the same shape, so it isn’t a fluke.

The case for meditation

Discomfort isn’t the same as harm. Sitting still without distraction lets whatever you’ve been avoiding come up. A survey asking runners about muscle pain, tiredness and low mood would find plenty, and nobody would call it injury. Buddhist traditions have names for these experiences (Tibetan nyams, for example), and a 2017 study from Britton’s lab found that they’re expected within the practice.

Two findings support this view:

• In Goldberg’s survey, 89% of meditators were glad they had practised, and the figure was essentially unchanged among those who’d had a bad effect.
• The only study here with a control group, Simonsson, Goldberg and colleagues (JMIR Mental Health, September 2026), randomly assigned 5,049 US and UK adults to a five- or ten-minute meditation recording, or to using the internet as usual. The meditators were no more likely to report a distressing experience than the people browsing the internet, and big jumps in bad mood were less likely after meditating.

When you add a control group, the side effect disappears — for ten minutes.

Dose

That trial tested one session, with questions asked straight afterwards and no follow-up, in people who mostly hadn’t meditated much. (Ten minutes carried higher odds of distress than five, but that was just as true for the people on the internet, so it says nothing about meditation in particular.) The surveys are where dose shows up. In Goldberg’s, more lifetime hours and more recent practice went with more bad effects. In Van Dam’s, more hours went with more impairment once other factors were controlled, and retreat-goers were more likely to report distress.

A short, single session is about as risky as scrolling. Years of practice, or a silent retreat, is a different thing, and those are the people the surveys are describing.

There’s also a warning about pushing through. Van Dam’s team caution against the idea that it has to get worse before it gets better. In Britton’s 2021 study of an eight-week mindfulness course, people who ended up with lasting bad effects were practising more on their own afterwards, about 13 times a week against about four. The study can’t say which came first.

Why the numbers disagree so much

The rate you find depends on how you look. Van Dam’s team note that trials which wait for people to report problems find bad effects in under 1%, while trials that actively ask find them in up to two-thirds. A 2020 review of 83 studies found 3.7% in experiments and 33.2% in observational studies of people who already meditate. In Britton’s study, an open question (“did anything bad happen?”) missed nearly 70% of what specific questions found.

Checklists can overshoot too: 97% of Van Dam’s sample reported at least one “unusual experience”, which says more about being human than about harm. The 60% is a checklist figure, the 9% is a real signal, and the 4% is the one worth worrying about.

Who it happens to

People reporting more psychological distress were more likely to report bad effects. That doesn’t mean distress causes them. Many people start meditating because they’re struggling, and the authors say plainly they can’t tell which came first.

Either way, the people most drawn to meditation are the ones who most need someone to check in. Van Dam’s team recommend telling people up front what might happen (“a critical part of informed consent”), considering mental-health history, and routine monitoring.

Like therapy, minus the therapist

Britton’s own comparison is the useful one. Lasting bad effects in her mindfulness course (6–14%) fell in the same range she cites for psychotherapy (3–14%). Nobody calls therapy dangerous, but therapy comes with an intake, check-ins and a person whose job is to notice.

In Van Dam’s sample about three in four had used a meditation app at least sometimes. App use wasn’t linked to more bad effects, so this isn’t a case against apps. The concern is that a practice with roughly therapy’s side-effect rate is mostly delivered by something that can’t notice anything. That’s an opinion, not a finding.

What helped people in the 2017 study wasn’t a clinician. It was having someone to talk to who understood.

---

Short is fine: the best evidence says five or ten minutes is about as risky as browsing the internet. If something comes up that frightens you, open your eyes. That isn’t failing. If it keeps coming back, tell someone: a teacher, a friend, your GP. In New Zealand, you can call or text 1737 any time, free.

---

Sources and reporting notes: all American or British samples, mostly online panels and self-report. Nothing from New Zealand that I found. Van Dam’s distress measure covers the 30 days before the survey, not before people started meditating; the press release says otherwise. The psychotherapy comparison is Britton’s. The Simonsson trial and the Van Dam paper were read via summaries of the full text.