Exercise, better sleep, a walk in parkland and a week off social media each lifted mood in trials

Four ordinary changes

Can anything ordinary move low mood, stress and brooding? Things a person can do this week, without a prescription or a waiting list. This article puts four sets of trials side by side that tested exactly that. None of them is about porn. They are about depression, sleep, a walk and social media, and they should be read that way.

The four changes are exercise, better sleep, a walk through parkland and a week away from social media. Each one moved mood in a randomised trial or in a review of such trials¹²³⁵. Each one also has a weak spot, and the weak spots matter as much as the headline numbers.

Exercise, against therapy and pills

The largest body of evidence is for exercise. A 2024 review in the BMJ by Noetel and colleagues pooled 218 trials with 14,170 people who had depression¹. It was a network meta-analysis, which can compare treatments that were never tested head to head through the control groups they share¹.

Against control groups such as usual care or a placebo pill, walking or jogging cut depression symptoms by an effect size of 0.63, in 1,210 people¹. Yoga came out at 0.55 and strength training at 0.49¹. Cognitive behaviour therapy on its own also scored 0.55, and SSRI antidepressants on their own scored 0.26¹. In this comparison a walk or a jog did about as well as talking therapy, and better than pills alone. The authors add that their review was not designed to find every trial of therapy or pills, so those two figures should not replace reviews built for them¹.

Harder exercise did more. Vigorous exercise such as running came out at 0.74, and light activity such as walking or hatha yoga at 0.58¹. Dance topped the table at 0.96, but that number rests on 107 people, 88% of them women with a mean age of 31, so we would not build a plan on it¹. Fewer people dropped out of strength training and yoga than out of the control groups¹. That counts for something. A treatment people quit does nothing for them.

Now the weak spot. Only one of the 218 trials met the Cochrane standard for low risk of bias¹. The authors rate their confidence as low for walking or jogging and very low for the other exercise types, and they did not grade therapy or pills alone¹. Few trials blinded participants or staff, so part of the effect may come from people expecting to feel better¹. A May 2024 correction says the effect sizes were standardised a different way from the one the methods section described, and that this did not materially change the findings¹. The fair reading is that exercise belongs in the same conversation as therapy, though the authors call the evidence for exercise less strong than for cognitive behaviour therapy¹.

Does better sleep come before better mood?

Sleep is the obvious next question, because it is hard to tell whether poor sleep drags mood down or low mood ruins sleep. A 2021 review by Scott and colleagues tried to settle it with trials².

It took in only trials where an intervention had improved sleep, most often cognitive behaviour therapy for insomnia². That let it ask a narrow question. When sleep got better, did mental health follow? Across 65 trials with 8,608 people, sleep quality improved by an effect size of 1.07 at the first follow-up². Depression fell by 0.63, anxiety by 0.51 and stress by 0.42². The more sleep improved, the more mental health improved².

The depression figure, 0.63, is the same as the one for walking or jogging in the exercise review¹². The two reviews looked at different people in different trials, so the match is a coincidence and not a ranking. It is still striking. Two ordinary changes, tested apart from each other, landed in the same place.

Rumination also fell, with an effect of 0.49, but only 4 comparisons with 355 people measured it, too few for firm conclusions, the authors say². Hold on to that for the next section. How the help was given mattered too. Sleep interventions given face to face by a clinician had an effect of 0.63 on overall mental health, against 0.34 for self-help ones².

The caveats here are real. The review found signs of publication bias, and adjusting for it brought the overall effect down from 0.53 to 0.35, though that was still statistically significant². Effects were also smaller at follow-ups of 6 months or more, 0.18 against 0.60 for shorter ones². Few trials measured anything beyond depression and anxiety². Better sleep was followed by better mood. How much, and for how long, is less certain than the headline numbers suggest.

Brooding, and where you walk

If brooding eases when sleep improves, can it be moved more directly? A Stanford study by Bratman and colleagues tried one of the simplest ways there is³. Sixty adults in California were sent at random on a 50-minute walk³. Half went through grassland with scattered oaks near the campus. Half went along a road with three to four lanes each way³.

After the nature walk, anxiety fell from 40.13 to 35.73 on a scale from 20 to 80³. Rumination fell from 38.93 to 36.43³. After the city walk, both barely moved, anxiety from 36.60 to 36.23 and rumination from 37.33 to 37.23³. Negative mood also fell more after the nature walk, and scores on one working memory test rose, though three other cognitive tests showed no difference between the walks³.

A second study by the same group sent 38 healthy city dwellers on a 90-minute walk⁴. Rumination fell after the nature walk and not after the city walk, and brain scans showed less activity in an area linked to rumination⁴. The difference in rumination between the two walks fell short of the usual test, though, with p at 0.07⁴.

This is the thinnest evidence in the article. The nature group in the first study started out more anxious, 40.13 against 36.60, which gave them more room to fall, though the authors report no significant difference between the groups at the start³. Everyone took one walk and was measured on return to the lab, and neither study tracked whether the benefit lasted³⁴. The volunteers were healthy, with no mental health diagnosis, and young, with a mean age of 22.9, which the authors list as a limitation³. What the studies do show is that the same walk in a different place did not do the same thing. If you are going to walk anyway, the route seems to matter.

Taking something away

The first three changes add something to the day. The last one removes something. In a trial at the University of Bath, published in 2022, Lambert and colleagues told 154 adults at random either to stop using Facebook, Instagram, Twitter and TikTok for a week or to carry on as usual⁵. They had a mean age of 28.9 and were not paid⁵.

The break group's own reports put their weekly time on those apps at 509.6 minutes before the trial and 20.7 minutes during it, about 8.5 hours down to 21 minutes⁵. The as-usual group went from 484.5 to 445.5⁵. Wellbeing in the break group rose from 46.00 to 55.93 on a scale from 14 to 70, against 43.92 to 45.05 in the other group⁵. The starting scores cover all 154 people and the end scores the 140 who finished⁵. Depression fell from 7.46 to 4.84 and anxiety from 5.95 to 3.88 in the break group, both more than in the group that carried on⁵.

Allowing for starting scores, age and gender, the break group ended 4.9 points higher on wellbeing, 2.17 lower on depression and 1.68 lower on anxiety⁵.

One week is short, and the weak spots follow from that. The trial cannot say whether the gains last⁵. Participants knew which group they were in and scored themselves⁵. That leaves room for hope to shape the answers. People who did not want a break chose not to sign up, so the sample may lean toward lighter or more motivated users, the authors note⁵. The trial also ran from November 2020 to March 2021, during the COVID-19 pandemic, and 62% of the participants were women, in a sample the authors describe as mostly young, white and in higher education⁵.

A larger US experiment points the same way with a stronger design. Allcott and colleagues recruited 2,743 Facebook users and, among the 1,661 willing to quit for less than $102, paid those picked at random to deactivate for four weeks ending just after the 2018 midterm elections⁶. Deactivating freed up about 60 minutes a day and raised wellbeing by 0.09 standard deviations⁶. That is a far smaller gain than the Bath trial found. A bigger, paid, month-long study finding less is a good reason to treat the one-week numbers as a ceiling rather than a promise.

How they found it

• The exercise review was a network meta-analysis, which can compare treatments that were never tested head to head through the control groups they share. Dance had the largest effect, 0.96, but from only 107 people.
• In the exercise review the effect grew with intensity. Vigorous exercise such as running came out at 0.74 and light activity such as walking or hatha yoga at 0.58.
• The sleep review only included trials where the sleep intervention worked, so it could ask whether better sleep was followed by better mental health. Most used cognitive behaviour therapy for insomnia.
• The nature walk went through grassland with scattered oaks near Stanford. The city walk followed a road with three to four lanes each way. In a second study, 38 healthy city dwellers took a 90-minute walk.
• In the social media trial, adults with a mean age of 28.9 were recruited through social media, word of mouth and local news, and were not paid. Self-reported weekly use fell from 509.6 minutes to 20.7 in the break group.
• A larger US experiment recruited 2,743 Facebook users and, among the 1,661 willing to quit for less than $102, paid those picked at random to deactivate for four weeks ending just after the 2018 midterm elections. Deactivating freed up about 60 minutes a day and raised wellbeing by 0.09 standard deviations.

What it cannot tell you

• Only one of the 218 exercise trials met the Cochrane standard for low risk of bias. The authors rate their confidence as low for walking or jogging and very low for the other exercise types. They did not grade therapy or pills alone.
• Few exercise trials blinded participants or staff, so expectations may account for part of the effect.
• The sleep review showed signs of publication bias. Adjusting for it brought the overall effect down from 0.53 to 0.35. Effects were smaller at follow-ups of 6 months or more.
• The nature group started out more anxious, 40.13 against 36.60, a gap the authors found not significant. One walk each, measured on return to the lab, and neither walking study tracked whether the benefit lasted.
• The social media trial ran for one week, during the COVID-19 pandemic. Participants knew which group they were in, and the scores were self-reported.
• Depression, sleep, walking and social media in general, not porn.

In P-Minus

On the P-Minus help button, the body step asks you to stand up, leave the room and walk for two minutes. Breathing and squeezing your muscles are the alternatives. P-Minus also asks what time you go to sleep, and reminds you at that time to thank God thirty-three times before you sleep.

Sources

1. Noetel M, Sanders T, Gallardo-Gómez D, Taylor P, del Pozo Cruz B, van den Hoek D, et al. Effect of exercise for depression: systematic review and network meta-analysis of randomised controlled trials. BMJ. 2024;384:e075847.
2. Scott AJ, Webb TL, Martyn-St James M, Rowse G, Weich S. Improving sleep quality leads to better mental health: a meta-analysis of randomised controlled trials. Sleep Med Rev. 2021;60:101556.
3. Bratman GN, Daily GC, Levy BJ, Gross JJ. The benefits of nature experience: improved affect and cognition. Landsc Urban Plan. 2015;138:41-50.
4. Bratman GN, Hamilton JP, Hahn KS, Daily GC, Gross JJ. Nature experience reduces rumination and subgenual prefrontal cortex activation. Proc Natl Acad Sci U S A. 2015;112(28):8567-72.
5. Lambert J, Barnstable G, Minter E, Cooper J, McEwan D. Taking a one-week break from social media improves well-being, depression, and anxiety: a randomized controlled trial. Cyberpsychol Behav Soc Netw. 2022;25(5):287-93.
6. Allcott H, Braghieri L, Eichmeyer S, Gentzkow M. The welfare effects of social media. Am Econ Rev. 2020;110(3):629-76.

Sources checked October 2026.