Why Meditation Makes Some People Feel Worse

Sit still. Close your eyes. Follow your breath.

For most people that instruction is harmless and often helpful. For a smaller group it is the beginning of a bad twenty minutes, and nobody warned them it might be. They come out of the session more wired than they went in, decide they are uniquely broken, and add one more item to the list of things that work for everyone except them.

That conclusion is wrong, and the mechanism behind it is not mysterious.

The evidence that this is real

The wellness conversation treats meditation as a substance with no side effect profile. The research literature does not.

A 2020 systematic review led by Miguel Farias at Coventry University pooled data across dozens of studies and found that adverse effects were reported by roughly 8 percent of participants. That places meditation in a similar range to psychotherapy in general, which also helps most people and harms a few. The most commonly reported effects were anxiety and depression, not exotic states.

Separately, a team at Brown University led by Jared Lindahl published a detailed qualitative study in PLOS ONE in 2017, interviewing experienced Western Buddhist meditators about difficult experiences. They documented a wide range of challenges, including anxiety, dissociation, emotional flooding, and changes in the sense of self. These were not beginners who quit after a week. Many were long-term practitioners and teachers.

Willoughby Britton, who ran that lab, has been blunt about the framing problem: the field spent decades studying benefits and comparatively little time studying who gets hurt.

None of this means meditation is dangerous. It means it is an intervention, and interventions have dose-response curves and contraindications. Treating it as universally safe is the error.

Mechanism one: you removed all the anchors

Start with what a silent seated practice actually does to your sensory environment.

You close your eyes, which removes vision. You sit still, which removes movement and most proprioceptive change. You go quiet, usually somewhere quiet. In the space of about thirty seconds you have stripped out nearly every stream of external information your nervous system normally uses to orient itself.

Orientation is not a luxury. A threat-detection system that cannot check the room has one remaining place to look, and that is inward.

So attention lands on the body. Heart rate, breath, gut, muscle tension, the small unexplained sensations that are always present and normally filtered out. For a system that is not currently scanning for danger, noticing all of that is neutral or pleasant. For a system that is scanning, it is a fresh supply of ambiguous signals to evaluate.

This is why the same instruction produces opposite results in two people sitting in the same room.

Mechanism two: attention is itself an input

There is a second problem layered on the first, and it is the one almost nobody names.

Watching a physiological process changes it.

Ask someone to observe their breathing and the breathing stops being automatic. It becomes a thing being performed, monitored, and evaluated. People report air hunger, the sense that they cannot get a full breath, or a compulsion to take a deliberate deep one. The deep one usually makes it worse, because over-breathing lowers carbon dioxide and that produces lightheadedness, which the system reads as another symptom worth worrying about.

The loop closes: you are anxious, so you monitor your body, and the monitoring generates sensations, and the sensations confirm that something is wrong.

For a person prone to panic, “focus on your breath” can be close to the worst available instruction, because breath is exactly the channel their alarm system is already watching.

Mechanism three: your starting state decides the dose

This is the piece that reframes everything else.

Most nervous system advice is written as though there is one direction to travel, and that direction is down. Calm down. Slow down. Settle. That framing assumes the problem is always too much activation.

It often is. It is not always.

A meaningful number of people are not revved up. They are flattened. Low energy, low motivation, low emotional range, the specific tiredness that sleep does not fix. That state needs the opposite input. It needs stimulation, movement, temperature change, light, engagement, something with an edge to it.

Give a flattened system a long silent sit in a dim room and you have prescribed more of what it already has too much of. People describe the result as heaviness, fog, a sense of sinking, or a flatness that lasts for hours afterward.

Same technique. Same instructions. Opposite starting state. Opposite outcome.

This is also, incidentally, one reason the research literature on nervous system interventions looks so inconsistent. Trials recruit for a diagnosis rather than for a physiological state, so participants who need opposite inputs get averaged together and the mean lands on nothing.

Mechanism four: stillness is not neutral for everyone

For people carrying trauma, immobility carries information.

Being unable to move is a feature of many overwhelming experiences. The body records it. Deliberately holding still, in silence, with eyes closed and attention turned inward, can reproduce enough of that configuration to trigger the associated state. Not as a memory exactly, more as a physiological echo that arrives before any thought does.

Some practitioners describe this as the practice “opening things too fast.” The mechanics are less mystical than that phrasing suggests. You have removed the external anchors that were doing the work of signaling safety, and you have asked the system to sit inside a posture it associates with helplessness.

This is why recommending silent meditation to trauma survivors as a first intervention is a genuinely questionable default, and why trauma-adapted approaches keep the eyes open, keep an external anchor available, and keep sessions short.

What to change before you abandon the whole category

If sitting practice consistently makes you feel worse, three variables are worth adjusting before you conclude that nothing works.

Keep your eyes open. Soft gaze, resting on a fixed point a few feet away. This preserves the visual anchor and the orienting information that comes with it. Most of the benefit that people attribute to closed-eye practice does not actually require closed eyes.

Anchor outside the body. Instead of the breath, use a sound in the room, the pressure of your feet on the floor, the temperature of the air on your hands, or a distant object. You get sustained attention training without turning your interoceptive alarm system into the object of study.

Cut the duration hard. Two minutes, not twenty. If two minutes is fine and ten is not, you have found a dose, not a failure. Build from the dose that works rather than from the dose someone put on an app.

And if your baseline is flat rather than activated, invert the whole prescription. Walk first. Get cold water on your face. Get outside into bright light. Do the settling work afterward, once there is something to settle.

The honest limits of this article

Meditation has a large and reasonably good evidence base for anxiety, depression and stress. The average effect is positive. Most people who practice it are fine, and a lot of them get real benefit.

This article is about the minority for whom the standard instruction backfires, and about why “try harder” is the wrong response when it does. Both things are true at once. A practice can be genuinely useful at the population level and genuinely wrong for a specific nervous system on a specific day.

If sitting practice reliably produces panic, dissociation, or intrusive memories, that is worth raising with a clinician rather than pushing through alone. Clinical care, a stable home, steady income, and real support are the floor. Everything written here is built on top of that floor and none of it is a substitute for it.

The useful question was never whether meditation works. It is which direction your system is stuck in, and whether the input you were handed moves you in that direction or the other one.