Desk stretches are the most recommended and least examined movements in fitness writing. Nearly every list of them is sold on a claim the evidence will not carry — that they correct your posture, or undo what sitting has done. This page keeps the movements and drops the claim, because the movements survive without it and the claim does not survive at all.
The honest case for leaving the chair is smaller than the usual one and much better supported.
The short answer
- They reduce the discomfort of having held one position. That effect is real, immediate and short-lived.
- They do not correct posture, and posture is a smaller lever than it is sold as.
- Frequency does the work. Two minutes several times a day beats twenty minutes once.
- The best-supported workplace intervention for office neck and shoulder pain is not stretching at all. It is a small dose of resistance training.
- Anything sharp, electric or numb is not a stretch. Ease out of it.
What desk stretches are for
Sitting is a held position. Tissue tolerates one for a while and then stops enjoying it, and the ache that follows is a signal about duration rather than a report of damage. Moving to the end of a range — any range, in any direction — changes the loading, moves blood, and resets the sensation. The relief is immediate and it wears off, which is the clearest evidence about what it is: a break from a position, not a repair to a body.
The workplace break literature supports roughly that much and no more. Waongenngarm and colleagues reviewed break interventions in office workers in Applied Ergonomics in 2018 and found small reductions in discomfort from more frequent short breaks, with productivity unharmed. Galinsky and colleagues, in the American Journal of Industrial Medicine in 2007, gave data-entry operators supplementary breaks with a short stretching routine in them and measured lower discomfort and no loss of output.
Grade: weak but consistent, and the intervention is nearly free. These are small studies with self-reported outcomes and short follow-ups. What makes them worth acting on is not the strength of the finding — it is that the cost of acting is two minutes and the cost of being wrong is nothing.
Eight that fit beside a chair
These are the desk-viable movements from Limber's own catalogue, with the app's default holds. The catalogue holds 56 moves today; 19 of them need no equipment at all, and these eight are the ones that read as ordinary office behaviour.
- Chin tucks. Draw the chin straight back over the throat, not down. Eight slow repetitions, one per breath. Dynamic, not a hold.
- Neck side tilt. Ear toward the shoulder, the other shoulder heavy. 30 seconds each side.
- Anchored upper-trap stretch. Sit on one hand, tilt the head away from it. 30 seconds each side. The anchoring is what makes it a stretch rather than a lean.
- Levator scapulae stretch. Turn the head about 45 degrees, then look down into the armpit. 30 seconds each side.
- Scalene stretch. Tilt away, then rotate the chin slightly up toward the ceiling. 30 seconds each side. Gentle — this one is easy to overdo.
- Seated tall rotation. Sit tall, cross the arms, turn. 40 seconds each side, held rather than bounced.
- Standing side bend. One arm overhead, reach long and lean. 30 seconds each side.
- Cross-body shoulder hold. Arm across the chest, held above the elbow. 30 seconds each side.
Two more are worth having if the room allows: a doorway pec stretch needs a doorframe, and wall slides need a clear wall. A wrist flexor stretch, 30 seconds a side, is the one most typing days actually ask for and the one nobody lists.
That is about seven minutes if you do everything. Nobody does everything. Pick two or three and repeat them across the day — see stretches at your desk with no equipment for the versions that survive an open-plan office, and the five-minute desk routine for a fixed order you can run without deciding anything.
The dose, and where it came from
The stretching dose literature is narrow but unusually consistent. Bandy and Irion, in Physical Therapy in 1994, compared 15, 30 and 60 seconds of static hamstring stretching over six weeks in 57 people: 30 seconds beat 15 and matched 60. Their 1997 follow-up, with 93 people, found one 30-second hold a day matched three.
Thomas and colleagues examined the wider question in the International Journal of Sports Medicine in 2018 and reported the pattern that has survived best: total accumulated stretching time across the week matters more than how you slice it up.
Grade: reasonably solid, and narrow. Young healthy adults, mostly hamstrings, six weeks or less. Nothing in it was measured on a neck at a desk. It is still the best guide anybody has, and its practical reading is that 30 seconds is enough and that more is not faster.
The strongest finding is strength
Here is the part that costs us something to print, on a page written by a mobility company.
The best-supported exercise intervention for neck and shoulder pain in office workers is not stretching. Andersen and colleagues, in Pain in 2011, randomised office workers with frequent neck and shoulder pain to a two-minute daily dose of progressive resistance training — a single lateral raise with a dumbbell — and measured a real reduction in pain over ten weeks. Their earlier trial, in Arthritis & Rheumatism in 2008, found specific strength training for the painful muscles outperformed general fitness training. Blangsted and colleagues ran a one-year randomised trial in the Scandinavian Journal of Work, Environment & Health in 2008 and found specific resistance training reduced neck and shoulder symptoms where an all-round activity programme did not.
Grade: moderate, and it is the best evidence in the whole desk literature. Randomised, workplace-based, with a real clinical outcome.
So the honest ordering is: if your neck and shoulders hurt at work, a very small amount of resistance training has better evidence behind it than any stretch on this page. Stretching is for the stiffness of a held position. Those are two different problems and the category runs them together.
What they will not do
They will not correct your posture. Richards and colleagues, in Physical Therapy in 2016, sorted the sitting neck postures of more than 1,000 adolescents into distinct clusters and found the clusters did not differ in neck pain the way the model predicts. Swain and colleagues, reviewing the systematic reviews in the Journal of Biomechanics in 2020, found no consensus that spinal posture causes low back pain.
State the counterweight at full strength, because it exists. Mahmoud and colleagues, in Current Reviews in Musculoskeletal Medicine in 2019, meta-analysed forward head posture and neck pain and did find an association in adults, though not in adolescents, with small to moderate effect sizes from cross-sectional studies. That is a real finding and it is the strongest thing the posture side has. It is also a correlation measured at one moment, which cannot tell you whether the posture came first. Forward head posture works through it properly.
They will not undo sitting. The mortality association with sitting time, pooled across more than a million adults by Ekelund and colleagues in The Lancet in 2016, was largely attenuated by 60 to 75 minutes a day of moderate activity. That is a walk, not a stretch. A desk break is a good thing that answers a different question.
And they will not tell you what is causing pain. Neck and back pain at a desk are common and usually settle. Pain that radiates down an arm or a leg, numbness, weakness, pain after a fall, or pain with fever or unexplained weight loss belong to a clinician who can examine you. Nothing here diagnoses anything and nothing here is treatment.
The area a desk hits hardest
The upper back is the part of you a screen day removes most completely: thoracic extension and rotation are the two movements that simply do not happen while typing.
And here is an admission about our own instrument. Limber's battery has five tests and none of them reads the thoracic spine. The reading for that area comes from a seated rotation test that is not built yet, in the app or on the web, so the area a desk affects most is the area our own score cannot see. The battery reaches four of the five score groups and the site says so on every page that composes one. The neck is a second gap of the same kind: there is no neck test, and the neck area is inferred from the shoulder reach at full weight — a proxy, named as one.
We would rather print that than pretend a fold-forward measured your upper back.
Take a reading, then work
Everything above is small in size, and small effects with strong feelings are exactly the combination that fools people. You cannot tell from how Thursday felt whether four weeks of breaks changed anything.
Take the shoulder mobility test — the Apley scratch, both sides, five bands — and the toe touch test for the posterior chain, write the date beside each, then take them again in three weeks with the same protocol at roughly the same hour. Limber's default re-test interval is 21 days, adjustable to 14 or 28, and the same interval works here.
In the app, the desk case is a built four-week track — twelve sessions, three a week, working the upper back, neck and shoulders from gentle to full extension — and the plan engine's default goal is desk relief. What it does not have is a reading for the area that track works hardest, for the reason given above.
The wider argument sits in desk and posture, and the break evidence is set out in full in the micro-breaks research.
Questions
Do desk stretches actually work?
They reliably reduce the discomfort of having held one position, and that effect is real, immediate and short-lived. They have not been shown to correct posture or to prevent pain. Treated as breaks from a position rather than repairs to a body, they are easy to justify and easy to keep doing.
How often should I do them?
More often than you do, and the exact interval matters less than the pattern. Break studies use intervals from every 20 minutes to every hour and find small benefits across that whole range with no clear optimum. A couple of minutes each hour is a defensible default because it is frequent enough to matter and light enough that you will actually do it.
How long should I hold each one?
Thirty seconds. Bandy and Irion found 30 seconds beat 15 and matched 60 over six weeks, and their follow-up found one hold a day matched three. Longer holds and extra repetitions have not earned their time in any trial, and at a desk the practical limit is what you will repeat rather than what is optimal.
Will stretching at my desk fix my posture?
No, and nothing has been shown to. Exercise programmes can shift measured posture angles a little, but nobody has shown that moving a person's habitual posture toward a textbook one reliably changes their symptoms. Posture is a much smaller lever than the advice implies, and a routine sold as correction is promising something it cannot measure.
What if my neck or back actually hurts?
Then the better-supported answer is a small daily dose of resistance training rather than stretching — that is what the randomised workplace trials show. And persistent pain, pain that radiates into an arm or a leg, numbness, weakness or pain that followed an injury is a matter for a clinician who can examine you, not for a longer hold.
Which stretch should I pick if I only do one?
A seated tall rotation, 40 seconds each side. A screen day removes thoracic rotation more completely than any other movement, it is the direction hardest to get back by accident, and the position looks like somebody turning to talk to a colleague.