Scapular Winging from Nerve Palsy

Face a wall, put both hands on it at shoulder height, and push. Have someone film your back.

If the inner edge of one shoulder blade peels away from your ribs, that points at one nerve. If instead the whole blade drops and slides outward and your shrug is weak, that points at a different one. That single observation decides the plan, so take the video to your appointment and ask specifically for a nerve conduction test.

Takes 15 seconds. No equipment beyond a phone.

Before you do anything else: if the weakness is getting worse rather than settling, if both sides do it, if anyone in your family has had the same thing, or if you have new headache or neck pain, see a doctor first. Those point to a different diagnosis, and strengthening will not touch it.

The Verdict

One nerve stopped working, so your shoulder blade lifts off your back. Load it, don't rest it.

Your shoulder blade is not bolted to your skeleton. It floats on your rib cage, held flat by muscle straps the way a rucksack is held against your back. Cut the wiring to one strap and it goes slack, so the blade peels away every time you push. The strap itself is undamaged and so is the joint, which is why resting it changes nothing. The wiring has to regrow, and that runs in months and years rather than weeks.

What this actually is: the nerve feeding one muscle has stopped working, so that muscle can no longer hold your shoulder blade against your ribs. The muscle and the joint are both fine.
What most people get wrong: protecting it. The best trial in this area gave one group a proper progressive strength program and the other a gentle one, and the gentle group did worse on pain, disability and strength.
Start here: get a nerve conduction test, because which of the two nerves it is changes the whole plan, and roughly one in five people with this sign turn out not to have a nerve problem at all.

Best for

People with a confirmed nerve palsy that is stable or slowly improving, who want to keep training around it rather than stop.

Skip if

Your weakness is getting worse, both sides are affected, you have a family history of the same problem, or you have new headache or neck pain. Get assessed first.

Want the full evidence? Keep scrolling

What Works

Dark cinematic study of shoulder girdle musculature under load

An unusual note before the list. This condition has two possible nerves behind it, and they do not have equal evidence. The trapezius nerve has seven randomised trials. The serratus nerve has none at all. The tiers below say which is which rather than blending them.

1. Supervised progressive resistance training STRONG

Applies to: spinal accessory (trapezius) palsy

Two randomised trials agreeing. Progressive resistance beat a lighter standard protocol on shoulder pain and disability (-9.6 points, 95% CI -16.4 to -4.5, P=.001), strength (+10.8 kg, P<.001) and endurance (McNeely 2008, n=52). A multicentre trial found active overhead reach improved by 26.6 degrees at 3 months (McGarvey 2015, n=59).

Dose: 3 sessions per week for 12 weeks, progressive in sets, repetitions and load. Expect measurable change by 12 weeks, and benefits largely held at 12 months in the people who kept training.

2. Conscious control of where the shoulder blade sits MODERATE

Applies to: spinal accessory (trapezius) palsy

One randomised trial with concealed allocation and blinded assessment. Adding deliberate shoulder blade positioning on top of strengthening improved overhead reach by 19 degrees (95% CI 10-29, P<.001) over strengthening alone (Chen 2021, n=36). Note that pain intensity did not improve in either group, so this buys movement and range rather than pain relief.

Exercise Prescription

Where these numbers come from: the sets and repetitions below are standard clinical practice, not figures lifted from a trial. What the research actually supports is the pattern. Three times a week, twelve weeks, and gradually harder as you get stronger.

Wall push-up plus3 × 10-12 · 3×/week

Stand facing a wall, hands at shoulder height, arms straight. Push your chest away so your shoulder blades spread apart, then let them come back together slowly. You should feel effort under the armpit and along the ribs.

Wall slide with a band3 × 8-10 · 3×/week

Loop a light band around both forearms and press outward against it. Rest your forearms on the wall and slide them slowly up. Stop at the height where the shoulder blade starts to wing badly.

Row3 × 10-12 · 3×/week

Pull a band or cable toward your ribs, leading with the elbow, squeezing the shoulder blade back and down.

Shrug3 × 10-12 · 3×/week

Only if your shoulder sits visibly dropped. Hold a weight in each hand and lift your shoulders straight up toward your ears, then lower slowly.

Supported overhead press2 × 8-10 · 3×/week

Press a light weight overhead with your back against a wall or bench, which does the job the weak muscle cannot do yet. Stop if you cannot control the weight on the way down.

Tier 2 and Tier 3 — the rest of the hierarchy

3. Honest explanation and expectation-setting MODERATE

No trial isolates this, but the prognosis data make it the only defensible substitute for the two-year timeline patients are otherwise handed. Naming the nerve, showing which movement exposes it, and saying plainly that this runs in months to years.

4. Structured, reassessment-driven progression MODERATE

A stage-based program guided by repeat assessment beat a fixed conventional protocol on shoulder scores at 3 and 6 months (Zeng 2026, n=98). Non-English publication, effect sizes read from the abstract only.

5. Exercise selection guided by muscle activation data EMERGING

Adding an outward band press to wall push-ups and wall slides raised activity in all the target muscles and lowered it in the chest (Kim 2023, n=30). Read this precisely: it is a laboratory measurement of which muscles switch on, not evidence that switching them on changes an outcome.

6. The same resistance approach applied to the serratus nerve EXTRAPOLATION

No trial exists. This is the trapezius result carried across to a different nerve, a different cause and a different population. It is reasonable, and it is reasoning rather than evidence, which is how it should be described to a patient.

What Doesn't Work

Treating it as a movement-pattern problem. True nerve winging is a different disease from the timing problem it resembles, not its severe end. There is nothing to re-time in a muscle that has no nerve supply.
Protecting the shoulder from load. The strongest randomised finding here is that progressive resistance beat the gentler protocol on pain, disability, strength and endurance. The instinctive response is the arm that did worse.
Bracing. A shoulder brace for this exists as a preliminary report from 2006 and has never been tested against anything.
Quoting a two-year recovery timeline. It does not survive longer follow-up, and it sets up a specific failure at month 25.
Waiting for a scan to make the diagnosis. The test that answers this is electrical, not radiological.

Red Flags

See someone now if any of these apply

Dark cinematic anatomical study of the shoulder girdle and scapula

Weakness getting worse rather than staying stable or slowly settling. The whole evidence base assumes a static or recovering nerve.
Both shoulder blades doing it, or weakness in the face, or a family history of the same problem. This is the pattern of a muscular dystrophy, which accounted for 5 of 128 cases in the largest series and will progress straight through any exercise program.
New headache, new neck pain, or a history of head and neck cancer. Tumours pressing on the nerve and tears in a neck artery have both first shown up this way.
Winging that appeared after surgery on your neck, armpit or chest. After armpit lymph node surgery every case in a 214-patient study was already visible at the one-month check.
Severe shoulder or arm pain happening right now, out of proportion to anything mechanical. That is the acute inflammatory phase and it needs proper pain relief.
Weakness spreading beyond the shoulder blade, or breathlessness, which suggests a wider nerve problem.

Refer to: neurology or a shoulder specialist with access to nerve testing in every case. Urgent imaging for the headache, neck pain and cancer-history presentations. There is no version of this condition that is managed well without a nerve conduction test.

Return to Training

This is one muscle, not your whole shoulder. Everything below shoulder height, all lower body work, all trunk and arm work carries on at full load throughout. What comes out is unsupported overhead pressing, dips and floor push-ups.

Nerve conduction test done, and the diagnosis is a nerve palsy rather than a dystrophy or a structural lesion
Shoulder blade holds flat against the ribs through full overhead reach with no weight
Strength on the affected side within one grade of the other side
3 sets of 10 supported overhead presses at a light load with the blade controlled and no winging on the way down
Overhead reach matching the other side within 10 degrees
No increase in winging or in fatigue 24 hours after a session
Stable or improving over 2 consecutive weeks, not fluctuating

For competitive lifters: the clock here is nerve regrowth, not tissue healing, and it does not respond to how motivated you are. Planning a return around a competition date will not work. Plan around repeat examination findings instead.

Conviction

MODERATE

Confidence is not uniform across this page, so here it is claim by claim rather than as a single number.

Which direction the blade wings tells you which nerve it is HIGH
An inflammatory nerve condition is the dominant cause of the serratus type HIGH
About one in five cases of this sign are not a nerve palsy at all HIGH
Progressive resistance beats lighter exercise in trapezius palsy MODERATE
The same applies to serratus palsy LOW extrapolation only
An individual recovery timeline can be given LOW
Any drug treatment helps the inflammatory type NO EVIDENCE
Any bedside test has known accuracy NO EVIDENCE

What would change my mind on "load it, don't rest it"

A randomised trial of at least 120 people with a confirmed isolated serratus nerve palsy of non-surgical cause, comparing a defined progressive resistance program against a defined low-load program, with blinded assessment of overhead reach and winging severity at 12 months. A clear win for resistance would move that recommendation from reasoning to evidence. A null result would show the trapezius finding does not transfer, and would make watchful waiting the default.

What would change my mind on "no timeline can be given"

A study following at least 200 consecutive people from the moment of diagnosis rather than from referral to a surgical centre, with a definition of recovery agreed in advance and checks at 6, 12, 24 and 60 months, split by cause. That is the study that would replace "somewhere between a fifth and three quarters of people" with something you could actually say to a patient. Its absence is the biggest hole in this field.

The Full Picture — Anatomy, Diagnosis & Evidence

What's Actually Going On

Dark cinematic anatomical rendering of the scapula against the rib cage

The shoulder blade has no bony joint with the rib cage. It is held there entirely by muscle, which means losing one muscle exposes the fact that the blade is not attached to anything. Two nerves matter and they produce opposite pictures.

The long thoracic nerve feeds the serratus anterior, the muscle that wraps around your ribs under the armpit. Its real job is rotating the shoulder blade upward so the socket can meet the arm bone as you reach overhead. Lose it and the inner border lifts away from the ribs, and overhead reach fails somewhere around shoulder height because the socket cannot be rotated up to meet the arm.

The spinal accessory nerve feeds the trapezius, the big diamond-shaped muscle that suspends the whole shoulder girdle. Lose it and the blade drops and slides outward, the shrug goes weak, and reaching out to the side is limited. You can usually see this one before the person moves.

What causes the nerve to fail matters more than most clinicians expect. The dominant non-traumatic cause of serratus palsy is an inflammatory nerve condition rather than something pressing on the nerve: it accounted for 61 of 70 cases in the largest series. That reframes the whole thing from a mechanical problem to be unloaded into an inflammatory event with a slow recovery. The dominant cause of trapezius palsy, by contrast, is surgical, mainly neck operations.

How to Identify It

Dark cinematic study of the posterior shoulder and scapular border

Wall push-up — serratus anterior and the long thoracic nerve No published accuracy data
Resisted forward reach or punch — serratus anterior No published accuracy data
Resisted shrug — trapezius and the spinal accessory nerve No published accuracy data
Watching the blade through active reach — trapezius palsy No published accuracy data
Needle EMG with nerve conduction studies of both nerves, both sides — the reference standard Confirmatory

Why every one of those says no data, and why that is the finding rather than a gap. Across 98 papers, not one bedside test for this condition has a published figure for how often it catches the problem or how often it correctly rules it out. The literature explains its own silence: every study uses the nerve test to decide who is in the study, and uses the physical examination to decide who gets sent for the nerve test. No study set up that way can produce those numbers. Do not quote accuracy figures for these tests, because none exist. The examination's job here is to establish that the blade is the problem, that a nerve is the reason, and which nerve to name on the referral.

The number worth carrying: in the largest series, 24 of 128 people with a winged shoulder blade did not have either nerve palsy. Five had a muscular dystrophy, six were doing it voluntarily, eleven had a bone or joint cause, and two were never explained.

The Debate

No clinical practice guideline exists for this condition as of August 2026. No major body has published one. Every disagreement below is between studies, not between a guideline and a study.

How long does recovery take?

Traditional teaching

Most people recover on their own within two years.

VS

Geurkink 2023, systematic review, 23 studies

Spontaneous recovery ranged from 21% to 78% across studies, followed up for a median of six years.

Follow the newer data. The two-year figure comes from small studies with short follow-up and no shared definition of what "recovered" means. That is a measurement artefact, not a biological disagreement, and the honest position is that no individual timeline can be given.

Protect it or load it?

The instinct, and the comparison arm in both trials

Gentle therapeutic exercise for a weakened shoulder girdle.

VS

McNeely 2008 (n=52) and McGarvey 2015 (n=59)

Progressive resistance won on pain and disability, strength, endurance, and overhead reach.

Follow the trials, but stay inside their population. Both were done in people whose trapezius nerve was injured during neck surgery. Carrying the result across to the serratus nerve is reasonable and it is still an inference.

Is there a drug for the inflammatory type?

Expert recommendation, review 2016

A short course of high-dose steroids early, while the pain is still present.

VS

Cochrane review 2009

Zero randomised trials of any treatment were found. Thirty papers of anecdote, only three with more than ten treated patients.

Neither over-rules the other, because there is nothing to over-rule. Present it as an unproven option a doctor may consider in the early painful phase, not as evidence-based treatment.

Honest Limitations

The evidence and the commoner problem do not overlap

What the research shows: seven randomised trials support rehabilitating this shoulder, and progressive resistance beats gentler work.

The real-world gap: every one of those trials was done in head and neck cancer survivors whose trapezius nerve was cut or bruised during surgery. The commoner presentation in a physical therapy clinic is the serratus nerve failing from an inflammatory condition, in a younger and otherwise healthy person. Six studies covering 234 shoulders reported non-surgical outcomes for that group and not one of them randomised anything.

What we do about it: apply the loading principle anyway, and say out loud that it is reasoning rather than proof for that nerve. Do not attach the trials' numbers to it.

"Conservative management" is not a treatment in this literature

What the research shows: 21% to 78% of people recover with conservative management.

The real-world gap: in those six studies, conservative management means the person was not operated on. What they did, how much and how often is simply not reported. So that number is a natural history statistic wearing a treatment label, and the six-year median follow-up means those recoveries happen on a timescale no patient will accept as reassurance.

What we do about it: never cite the recovery range as evidence that a program works. Cite it as what happens over years, with or without you.

Supervised trial adherence is the ceiling, not the expectation

What the research shows: progressive resistance worked over 12 weeks.

The real-world gap: those trials achieved 93% to 95% attendance at supervised sessions inside a monitored hospital rehabilitation service. At 12 months the people still doing better were the ones who had carried on training, which is just as consistent with the motivated ones continuing as it is with the training causing the benefit.

What we do about it: build the program around what someone will actually do without supervision, and treat continuing to train as the intervention rather than the 12-week block.

The Nuance

Dark cinematic anatomical study of the shoulder girdle and surrounding structures

Surgery versus carrying on conservatively, with the numbers and the catch.

Reported spontaneous recovery without surgery is 21% to 78% across six studies, followed for a median of six years. For surgery, pooled results look considerably better: tendon transfer for serratus palsy improved overhead reach by an average of 47 degrees, cut pain scores substantially, and moved shoulder function scores a long way. Nerve release in 29 people operated a median of 30 months after onset significantly improved reach and reduced winging severity.

Here is the catch, and it is the whole point of this section. Every surgical number in this field comes from a series with no comparison group, where the surgeon assessed the outcome. The systematic review that pooled them found no high-quality study among the 23 it included. The comparison for any of those series is the same 21% to 78% who recover anyway, and nobody has ever randomly assigned people to surgery versus carrying on. So the surgical results look better than they can be shown to be.

What can be said fairly: a substantial share of people recover without surgery, recovery is slow enough that they need telling so at the start, and surgery is a reasonable conversation for someone still disabled at a year rather than something to be steered toward early. The usual trigger point in the literature is 9 to 12 months of conservative management first.

One more nuance about measuring progress. None of the standard shoulder questionnaires has a known threshold for meaningful change in this specific condition. Someone can improve their score by getting better at compensating while the paralysis itself is unchanged. Photograph the winging in a standard position and track that alongside the questionnaire.

Sources

Geurkink et al., 2023, Journal of Shoulder and Elbow Surgery. Systematic review with meta-analysis, 23 studies. Spontaneous recovery 21-78% at a median 72 months; tendon transfer improved forward flexion by 47 degrees. The only on-topic synthesis in the field
McNeely et al., 2008, Cancer. Randomised trial, n=52. Progressive resistance beat standard therapeutic exercise on shoulder pain and disability (P=.001) and strength (P<.001). Strongest interventional evidence here
McGarvey et al., 2015, Head & Neck. Multicentre randomised trial, n=59. Progressive scapular strengthening improved active abduction by 26.6 degrees at 3 months (p=.007).
Seror et al., 2018, Muscle & Nerve. Clinical and electrical series, n=128. Long thoracic palsy 70, spinal accessory 39, muscular dystrophy 5, voluntary 6, bone or joint causes 11. Inflammatory nerve disease caused 61 of the 70.
Seror et al., 2017, Muscle & Nerve. Trapezius palsy series, n=54. Inflammatory nerve disease the commonest medical cause; neck surgery the commonest surgical one.
van Alfen et al., 2009, Cochrane Database of Systematic Reviews. Zero randomised trials of any treatment for the inflammatory nerve condition were found. An evidence vacuum, not a null result
van Eijk et al., 2016, Muscle & Nerve. Review. Reported incidence of the underlying inflammatory nerve condition is 1 per 1,000 people, and it is routinely missed.
Wu & Ng, 2023, Journal of Hand Surgery Global Online. Review challenging the traditional two-year recovery teaching and setting the 9 to 12 month window before surgery is considered.
Ng et al., 2024, Journal of Hand and Microsurgery. Surgical series, n=29. Nerve release at a median 30 months improved reach and winging severity. Uncontrolled.
Chen et al., 2021, Integrative Cancer Therapies. Randomised trial, n=36. Adding conscious shoulder blade positioning improved abduction by 19 degrees over strengthening alone.

Go Deeper

Most injury advice tells you to rest something that needs loading, and gives you a recovery date nobody can actually predict. The Verdict digs into one condition a week and tells you where the evidence genuinely runs out.

Join The Verdict — free

One nerve stopped working, so your shoulder blade lifts off your back.