Meditation research has expanded enormously over three decades. The volume of publication is not the same as the strength of evidence, and this is a field where the gap between the two is unusually wide.
It's worth going through what holds up, because there's something real here and the overclaiming makes it harder to see.
The methodological problem
Several issues affect this literature and they need stating first.
Blinding is impossible. Participants know whether they're meditating. Expectancy effects are therefore uncontrolled, and expectancy is a substantial contributor to outcomes on self-reported measures.
Control conditions are frequently weak. Comparing meditation against a waiting list tells you that doing something beats doing nothing, which is not the question. Active control conditions — matched for time, attention, group contact and expectation — produce considerably smaller effects.
Small samples. Many studies enrol a few dozen participants, which produces unstable results and inflated effect sizes when significant.
Publication bias. Analyses examining this literature have found evidence that null results are underpublished, which inflates pooled estimates.
Self-selection. People who volunteer for meditation trials expect it to help, which is not a neutral starting point.
A major review commissioned to assess this field concluded that evidence quality was mixed, with moderate evidence for improvement in anxiety, depression and pain, and low or insufficient evidence for many other claimed outcomes.
Where the evidence is reasonable
Anxiety and depressive symptoms. Moderate evidence of small to moderate improvements, comparable to other active interventions. Not superior to established treatments in most comparisons.
Chronic pain. Some evidence for improvement, generally in the affective dimension — how distressing pain is — rather than in pain intensity itself. That distinction matters and is frequently collapsed in reporting.
Relapse prevention in recurrent depression. Mindfulness-based cognitive therapy has a reasonably robust evidence base for preventing relapse in people with multiple previous episodes, and appears in several clinical guidelines. This is arguably the strongest clinical finding in the field.
Stress in non-clinical populations. Reductions in self-reported stress are consistent, with the caveat about expectancy above.
Where claims outrun evidence
Structural brain changes. Widely reported and considerably more uncertain than coverage suggests. Early studies were small and cross-sectional, comparing long-term meditators with non-meditators, which cannot distinguish cause from selection. Longitudinal studies have produced inconsistent results and some have failed to replicate earlier findings.
Attention and cognition. Some evidence of improvement on specific attention tasks, with inconsistent replication and unclear transfer to everyday functioning.
Immune function and inflammation. Studies exist, effect sizes are small, and clinical significance is unclear.
Telomeres and ageing. Frequently cited, based on limited data, and best treated as preliminary.
The specificity question
The deeper issue underlying much of this. Does meditation work through the mechanisms its proponents describe, or through non-specific factors — expectation, relaxation, structured time, group support, the sense of doing something helpful?
Studies using well-matched active controls tend to find substantially smaller differences than those using passive controls, which suggests non-specific factors account for a meaningful share of the effect.
That's not a dismissal. Non-specific factors are real effects, and an intervention that reliably produces them is useful. It does mean claims about specific mechanisms should be held loosely.
The different practices problem
Meditation isn't one thing, and treating it as one confounds everything.
Focused attention practices, where attention is repeatedly returned to an object. Open monitoring, observing experience without a fixed focus. Loving-kindness practices cultivating specific emotional states. Body scanning. Mantra repetition. Movement-based practices.
These involve different cognitive operations and plausibly different effects. Pooling them and reporting on "meditation" produces an average across genuinely different interventions.
Duration and intensity also vary enormously between studies, from a few minutes daily over weeks to intensive retreat conditions.
The reasonable summary
Meditation appears to produce modest improvements in stress, anxiety and mood, comparable to other active interventions, with the effect partly attributable to non-specific factors.
It has a specific and reasonably well-supported clinical application in preventing depressive relapse.
It is inexpensive, portable, and low risk for most people, which makes the practical calculation favourable even given modest effects.
Claims about transformative effects, structural brain changes and physical health outcomes should be treated with caution.
And the honest framing for anyone starting: this is a practice that may modestly improve how you relate to your own mental states over months of consistent effort. That's genuinely worth something and it's a considerably more sustainable expectation than what the marketing offers.
A note on what practitioners say
Contemplative traditions generally didn't claim meditation was a stress reduction technique. They described it as a method for investigating the nature of mind, within a framework involving ethics and worldview.
Whether extracting the technique from that context preserves what's valuable is a genuine question, and one the research is not designed to answer. It's worth knowing that the thing being studied and the thing traditions transmitted are not obviously the same activity.