Can Art Reduce Stress? What Science Says

What changed in the research

For most of the twentieth century, studies of art and wellbeing asked people how they felt. Questionnaires, mood scales, self-report. Useful up to a point, though everyone can see what a survey handed out in an art gallery is fishing for.

The last two decades moved to measures you can't flatter. Salivary cortisol. Inflammatory cytokines. Heart rate variability. Skin temperature. Length of hospital stay. Fourteen-year mortality data. You can tell a questionnaire whatever you'd like to be true; your endocrine system isn't listening.

That shift is what makes the current evidence worth taking seriously. It's also what makes its limits so easy to see.

Twenty-two per cent, and why I'm only half convinced

Between July and September 2025, a team from King's College London led by Tony Woods took fifty people aged 18 to 40 into The Courtauld Gallery. They spent time with Manet, Van Gogh, Gauguin and Toulouse-Lautrec: the *Self-Portrait with Bandaged Ear*, *A Bar at the Folies-Bergère*. A control group looked at high-quality reproductions of the same paintings somewhere that wasn't a gallery.

Everyone wore research-grade sensors tracking heart rate variability and skin temperature. Saliva before, saliva after.

Cortisol in the gallery group fell by roughly 22 per cent. In the reproduction group, 8 per cent. The inflammatory markers IL-6 and TNF-α came down about 30 per cent in the gallery visitors and didn't move at all in the reproduction group. Skin temperature dropped 0.74°C and heart rate variability rose, which is roughly what a parasympathetic shift looks like from the outside.

Now my problem with it, which I'd rather say myself than have you find in a comment thread. The study was commissioned by Art Fund, an organisation whose entire purpose is getting people into galleries. It hasn't finished peer review. Fifty participants is a pilot. And nobody can be blinded to whether they're standing in the Courtauld or a seminar room, which means expectation is baked into the result with no clean way to subtract it.

What I'll take from it is the direction, not the decimal places. Something about the original object in the room appears to do work that a printout doesn't. If you've ever seen a painting in person after years of knowing it from screens, you already suspected as much.

The skill thing

Girija Kaimal's 2016 study at Drexel is the one everybody cites and almost nobody reads properly.

Thirty-nine healthy adults, a room with markers, paper, modelling clay and collage materials. No instructions of any kind: make whatever you want, forty-five minutes. Saliva before and after. Cortisol fell significantly, and 75 per cent of participants finished lower than they started.

The headline everywhere was "art reduces stress." The genuinely interesting result got buried: prior art experience predicted nothing. Trained painters didn't out-relax people who hadn't picked up a marker since primary school.

Having spent years around both artists and anxious people, I find that surprising. Skill normally changes everything about how someone relates to materials. Apparently not here. Whatever's doing the work isn't aesthetic competence but something duller and more useful: sustained manual attention on a task with no correct answer.

One caveat that matters. This was quasi-experimental, with no true control group, and cortisol drifts downward across the afternoon regardless. Some proportion of that 75 per cent is just time passing.

The bit about the brain that gets reported badly

Semir Zeki at UCL spent years putting people in scanners and showing them paintings. The finding that made the papers: looking at an image you personally find beautiful activates the medial orbitofrontal cortex, a reward region, with blood flow increases reported around 10 per cent, and activity in the caudate nucleus tracks how beautiful the viewer rates the work. The caudate also turns up in studies of romantic love, which is why every write-up of Zeki ends with a line about art being like falling in love.

Fine. But the result I keep coming back to is Edward Vessel's, and it barely gets mentioned anywhere.

Vessel and colleagues found that the artworks people rated as *most moving* didn't simply push harder on visual and reward circuits. They recruited the default mode network: the system associated with self-reflection, autobiographical memory, inward attention. That network is supposed to quieten when you focus on something external. In front of deeply affecting art it does the opposite.

If that holds, it changes what we think art offers a stressed person. It isn't distraction. Distraction pulls attention away from the self and the relief stops the moment the distraction does. This looks like the reverse manoeuvre: the work gives you somewhere to put yourself, and you come back slightly rearranged. Which would explain why a good hour in a gallery has a different half-life than an hour of television.

A window with trees in it

The most important study in this field contains no art whatsoever.

In 1984 Roger Ulrich published a single page in *Science*. He had gone through a suburban Pennsylvania hospital's records covering 1972 to 1981 and found twenty-three patients recovering from gallbladder surgery in rooms whose window faced a stand of trees, and twenty-three matched patients whose window faced a brick wall.

The tree patients went home sooner. They needed weaker painkillers, aspirin and paracetamol where the wall patients were on narcotics. Nurses wrote fewer negative comments about them in the notes.

Nothing was administered. The only variable was what sat outside the glass.

I raise it because it does something no art study has managed since. It establishes that visual environment is a physiological input rather than decoration, in a setting where the outcome measure was opioid consumption and nobody was trying to prove anything about culture. Once you've accepted that a view of trees alters a morphine requirement, "a painting alters cortisol" stops sounding like a wellness claim and starts sounding like plumbing.

Prescriptions

On 1 November 2018 the Montreal Museum of Fine Arts and Médecins francophones du Canada did something nobody had done before: they let doctors write prescriptions for museum visits. Up to fifty per physician per year, each admitting two adults and up to two children, free.

It reads like a stunt until you look at the target list. Eating disorders. Cardiac arrhythmia. Epilepsy. Alzheimer's. Breast cancer. Long-term mental health conditions. Nobody involved imagined Chagall was a treatment. The point was that patients need somewhere to be that isn't a waiting room and doesn't demand anything of them, ideally alongside another person, and that this counts as a clinical resource even though it doesn't look like one.

The idea has since turned into infrastructure. England's NHS runs social prescribing nationally, with something like 3,600 link workers connecting patients to arts, nature and community activity; around 1.3 million people had been referred by 2023. And in 2019 the WHO's European office published the field's most comprehensive review to date, mapping over 900 publications, including 200 reviews that between them covered more than 3,000 studies. Its conclusion was cautious and broad: the arts have a demonstrable role in prevention, health promotion and the management of illness across the lifespan.

Cautious and broad is usually what real evidence sounds like.

The mandala study, which is funnier than it sounds

In 2005 Curry and Kasser deliberately made 84 undergraduates anxious, then handed them one of three things: a mandala to colour in, a plaid pattern to colour in, or a blank sheet of paper to colour on freely.

Everyone expected the mandala to win. It arrives with a few thousand years of contemplative tradition and a marketing department's dream of a backstory.

The mandala and the plaid finished level. Both beat the blank page.

The sacred geometry was contributing nothing. The structure was contributing everything, and tartan works fine.

This is the single most practically useful result in the whole literature, because the wellness industry has the advice backwards. If you're already anxious and somebody tells you to sit with a blank canvas and express your feelings, they have handed you another demand at the exact moment you have no capacity for one. An empty page asks you a question. A pattern gives your hands somewhere to go. When you're depleted, take the pattern.

 Against calm

Here's where I argue with the title of my own article.

"Art reduces stress" smuggles in an assumption: that the best art is the most soothing art, and that lowered arousal is the thing worth measuring. As a theory of stress that's incomplete. As a theory of art it's dismal.

Stand in front of Munch's *The Scream*, or Goya's black paintings, or one of Doris Salcedo's installations, and nothing about your heart rate is going to improve. What happens instead is harder to chart. Something you have been carrying privately turns up outside you, made by a stranger, given a shape, survivable enough that a museum hung it on a wall. Psychology calls the mechanism aesthetic distance. Artists have been calling it the entire point for about three thousand years.

Zeki's own data is consistent with this, incidentally. Reward circuitry responds to work people find beautiful even when the content is grief. Being moved and being calmed are separate physiological events, and I see no reason to rank them.

The honest version of the claim is therefore smaller and more interesting than the headline. Certain kinds of engagement with art, held long enough, shift the body towards lower arousal. Other kinds do something else entirely, and a decent life probably wants both.

 Thirty-one per cent, with an asterisk

The *BMJ* runs a Christmas issue every year for research that's rigorous but slightly strange. In 2019 it carried a paper by Daisy Fancourt and Andrew Steptoe under a title that belongs on a novel: *The art of life and death.*

They followed 6,710 adults aged 50 and over from the English Longitudinal Study of Ageing for fourteen years, recording how often they went to museums, galleries, exhibitions, the theatre, concerts, the opera.

People who went every few months or more had a 31 per cent lower risk of dying during follow-up than people who never went. That's after adjusting for wealth, education, existing conditions, mobility, smoking, exercise and degree of social contact.

The asterisk is enormous and I won't bury it. This is observational. Statistical adjustment isn't randomisation, and the sort of person who can casually get to a concert on a Tuesday evening differs from the sort of person who can't in ways no model captures completely. Nobody is claiming a Rembrandt buys you time.

What's worth noticing is the shape of the finding. The protective signal attaches to frequency, not intensity. Every few months, sustained across fourteen years. Whatever this is, one overwhelming afternoon at the Rijksmuseum won't produce it. Having a standing relationship with art might.

 Where all of this falls apart

I've flagged individual problems as I went, but the structural ones deserve to sit together.

The samples are tiny. Thirty-nine people. Fifty people. Twenty-three surgical patients from the 1970s. These are pilot studies that got treated as verdicts largely because their conclusions were pleasant.

Cortisol is a noisy and badly over-interpreted measure. It swings with time of day, caffeine, sleep, and the mild stress of participating in an experiment at all, which is precisely why the missing control groups matter as much as they do.

And there's no such thing as a placebo Van Gogh. Blinding is unavailable in principle, so every result in this field contains an unknown quantity of expectation, framing, and the plain relief of having been granted permission to do nothing useful for an hour.

What survives all that is still worth something. Endocrine, immune, cardiac, neuroimaging, epidemiological and architectural methods, run largely by people who weren't talking to each other, have pointed the same direction for forty years. That isn't proof. It's enough to act on, which is a lower and more ordinary standard, and the one most of us live by.

 What I'd actually tell you to do

Take the evidence at face value and the practical advice looks nothing like "go to more museums."

**Go to originals when you can.** The cleanest signal in the recent data is the gap between the real object and the reproduction, and it isn't close. A screen is a different input.

**Give it forty minutes.** Forty-five keeps recurring across both the making studies and the gallery studies, with some suggestion that returns flatten after that. Twenty minutes is worth having. Seventeen seconds is a photograph.

**Pick one painting and stay with it for ten.** This is what the Slow Art movement built itself around and it's genuinely uncomfortable the first time. You'll be bored at ninety seconds and certain you've extracted everything by two minutes. The interesting part is on the far side of that.

**Make something badly with your hands.** Prior skill predicted nothing in Kaimal's data, so there is no version of this you're too incompetent for. Clay, markers, collage. Forty-five minutes, no brief, no photograph at the end.

**When you're depleted, take the structured option.** Pattern over blank page, every time.

And stop optimising for the transcendent afternoon. The mortality signal in the ELSA cohort attached to "every few months, for years," not to pilgrimages.

None of this asks you to believe anything mystical. It asks you to accept one fairly boring proposition: that sustained visual attention to a made object, in a room built for nothing else, is a physiological event with consequences.

Ulrich's patients never knew they were in a study. They had a window with some trees in it, and they went home two days early.

Whether you are looking to decorate one room or your entire house, or add professional credibility to your work place, you will find the perfect piece of art on Benarto’s Online Art Gallery. Click here to view their work.