A new study asks a narrow, careful question: after a month of OM meditation, does anything measurable change in how the brain handles a demanding task?

The answer is yes, a little, and only in one place. That modesty is what makes it worth writing about. Most of the work in reading a paper like this is not deciding whether to believe it. It is separating what was measured from what the measurement gets taken to mean — which turns out to be the central difficulty in this entire field.

I made a video on this, which covers the same ground more briefly.

The study, in one paragraph

Singh and colleagues, publishing in Applied Psychophysiology and Biofeedback in 2026, took 34 healthy adult men and split them into two groups: 18 practised OM meditation for a month, 16 did not. Before and after, everyone did a visual n-back task at three difficulty levels — 0-back, 1-back and 2-back — while their EEG was recorded. The researchers then looked at five event-related potential components to see whether anything had shifted.

That is a clean design. The question is what came out of it.

What moved, and what did not

Singh and colleagues (2026). Responses got faster at the two easier levels. The N2 shifted at the moderate level only. Accuracy was high everywhere, in both groups, which means the task was not straining anyone to their limit.

Three things are worth holding onto here.

Responses got faster, but only where the task was easy or moderate. At the hardest level, the meditation group was no quicker than the controls.

The brain-level effect is narrower still. It appears in one cell out of six.

And accuracy stayed high across the board. That matters more than it looks, and I will come back to it.

Faster is not the same as more

The headline most people would write from this is that meditation improved working memory. The data do not support that sentence.

Reaction time measures how quickly you produce an answer. Working memory capacity is how much you can hold and manipulate. These come apart routinely. You can get faster because you are more alert, because you have settled into the task, because you have adopted a slightly less cautious threshold for responding — none of which means you are holding more in mind.

If meditation had increased capacity, the obvious place to see it is the 2-back condition, where capacity is actually under pressure. That is precisely where the advantage disappears.

The high accuracy makes this sharper. When nearly everyone is getting nearly everything right, the task is not pushing against the ceiling of anyone’s capacity, so it cannot tell you much about where that ceiling is. Speeding up on a task you were already passing is a real result. It is not a bigger mind.

What an N2 can and cannot tell you

The N2 is a negative deflection in the EEG recorded a couple of hundred milliseconds after a stimulus, strongest over the front and middle of the scalp. It gets described as the brain’s conflict-monitoring signal.

That description is too tidy. Folstein and Van Petten reviewed this component for Psychophysiology in 2008, and the review’s own title gives the game away: cognitive control and mismatch. The N2 is a family of overlapping effects, not a single readout of a single process. What a given N2 shift means depends on the task, where on the scalp it sits, when it occurs, and what the behaviour did alongside it.

So “the N2 changed” is a measurement. “Cognitive control improved” is an interpretation with several steps between it and the measurement, and each step can fail.

There is a second limit worth naming. EEG is superb at when — it resolves events to the millisecond. It is poor at where. Scalp electrodes record the summed activity of large neural populations, and working backwards from the scalp to the sources is a mathematically underdetermined problem with no unique solution. A fronto-central N2 is consistent with medial frontal involvement, including the anterior cingulate. Consistent with is not the same as demonstrated, and scalp EEG on its own cannot close that gap.

One component out of five

This is the part I would most want addressed, and it is the one I nearly missed.

The researchers examined five ERP components — P1, N1, N2, P3 and a late slow wave. One of them produced a reliable effect. The other four did not.

Both readings are legitimate. Which one you should prefer depends on whether the N2 at moderate load was named as the prediction in advance, or found by looking.

I want to be fair about this. A specific, isolated effect is often what a real mechanism looks like — a genuine change in one process should not light up every component. And the researchers used cluster-based permutation testing, which is designed precisely to control for searching across many time points.

But the generous and the sceptical readings are only distinguishable by one thing: whether the N2 under moderate load was the stated hypothesis before the data were collected, or the result that survived a look across the grid. That is what a pre-registration answers, and it is why I now look for one first.

The comparison I would change first

Of everything here, this is the limitation that matters most, and it is not special to this study.

The meditation group was compared against a group that rested. Resting controls for the passage of time and for doing the task twice. It does not control for knowing you are in the meditation group, for expecting to improve, for the structure of showing up five days a week, or for the attention of whoever is running the programme.

Those are not small. They are exactly the ingredients that produce measurable improvement in trials of almost anything.

The field knows this. Van Dam and fourteen colleagues laid it out in Perspectives on Psychological Science in 2018, in a paper that remains the best thing written on why meditation is hard to study. Their figure is striking: only 9 per cent of mindfulness-intervention research had reached the stage of testing against an active control, against 20 per cent compared only with a waiting list or treatment as usual.

And when the comparison is made properly, the picture changes. Goyal and colleagues reviewed 47 randomised trials covering 3,515 people for JAMA Internal Medicine in 2014. Against non-specific controls, meditation programmes produced moderate benefits for anxiety, depression and pain — effect sizes around 0.3 to 0.4. Against specific active treatments — drugs, exercise, other behavioural therapies — they found no evidence that meditation was better.

So the single highest-value change to this study is not more participants. It is a control group that sits down for twenty minutes, five days a week, with someone paying attention to them, and does something that is not OM meditation.

Thirty days is a snapshot

The intervention ran for a month. That is enough to show that something shifted. It is not enough to show that anything lasts.

There were no follow-up measurements. So the results are consistent with a durable adaptation, and equally consistent with a temporary change in alertness or task strategy that fades within a fortnight of stopping. Nothing in the data distinguishes these, and the honest position is that nobody yet knows which it is.

What I would want is the boring, expensive thing: measure again at three months and at six, and measure again after people stop.

Thirty-four men

The sample was 34 healthy adult males — 18 and 16 in the two groups.

For the kind of effect being chased, that is small, and small samples do not just make it harder to find real effects. They also make any effect you do find more likely to be an overestimate, because only the larger fluctuations clear the significance threshold.

The all-male sample is a separate problem. It restricts the conclusions to men, and it rules out exactly the comparisons worth making — whether the effect differs by sex, age, health status, or how much someone had meditated before.

So how do you measure meditation?

Carefully, and with a clear head about what each instrument actually reports.

The reason this is hard is not that meditation is mystical. It is that almost every measurement here is a proxy standing in for something you actually care about, and the gap between proxy and target is where the overclaiming happens. Reaction time stands in for cognitive efficiency. The N2 stands in for cognitive control. A month of practice stands in for a changed mind.

Each of those substitutions is reasonable. Stack three of them and you can travel from “responses were slightly faster at 1-back” to “meditation rewires your brain” without any single step looking unreasonable.

What I take from this paper is not a conclusion about meditation. It is a demonstration of what a modest, well-bounded result looks like before anyone has stretched it: an effect that shows up at moderate difficulty and not at high, in one component out of five, in thirty-four men, over one month, against a group that did nothing.

That is a real finding. It is also almost exactly as small as it sounds, and the authors say so.

The interesting question it leaves is the one I would most like answered: whether meditation changes what you can do, or changes how you allocate what you already have. Those are different claims, they need different experiments, and this study gently points at the second.