Shoulders after sitting
The abstracts do not name a best gym lift. They name three kinds of work — heavy resistance, local stretches, and a stretch-plus-scapular package for desk angles — and they disagree about the spine.
Sit long enough and the shoulders come forward. Office papers enroll on that picture — protraction, a forward head, a rounded back, a cluster some of them call upper-crossed syndrome — and then they try to move the angles. The question that opened this thread wanted a lift list. One academic pass later, the wiki still does not have one.
What it has is kinds. Progressive resistance training beats doing nothing. Sleeper stretch and cross-body stretch move internal rotation. At the desk, stretching the pecs without training the scapula is the weaker half of the pair. The question page stays open: kinds, not a program, and not a diagnosis.
The lift list that isn’t there
The strongest strength evidence in the pass is not a shoulder paper. Currier’s 2026 ACSM overview — 137 systematic reviews, more than 30,000 people — says resistance training versus no exercise improves strength, size, power, endurance, and several function scores. Heavier loads, a full range of motion, two or three sets, the hard lift first, at least two sessions a week. Hypertrophy wanted at least ten sets a week and some eccentric overload.
The 2023 Bayesian network meta-analysis underneath it ranked a higher-load, multiset, thrice-weekly prescription first for strength, with a standardized mean difference of 1.60 versus control. Twice-weekly, same load idea, ranked first for size, at 0.66. Those reviews synthesize whole-body training. The retrieved abstracts do not name overhead press, lateral raise, seated row, or face pull as shoulder-specific winners. That is a hole, not a ranking of zero.
What actually lit up
Where the pass does name drills, it names activation, not a measured one-rep max. Reinold’s 2004 intramuscular EMG study, ten healthy adults, seven external-rotation drills: sidelying external rotation at zero abduction produced the highest infraspinatus and teres minor readings — 62 and 67 percent of a maximum voluntary isometric contraction. Prone horizontal abduction at 100 degrees with the arm fully rotated out lit the supraspinatus and both heads of the deltoid — 82, 87, and 88 percent. Reinold framed the work as information for people writing rehab programs, and for later studies. It is not a treatment card.
Reinold 2004 · peak EMG
Healthy adults, intramuscular EMG. Activation, not a 1RM ranking. Numbers from the shoulder-exercises wiki’s external-rotation page.
A later, smaller study asked a different question. Vetter, 2023: sixteen active men, twice-weekly eccentric isokinetic external rotation for about six weeks. Eccentric strength rose 24 percent. Supraspinatus fascicle length rose 16 percent. Passive range on a stretch test fell 4 percent. That is adaptation on a dynamometer, not EMG, and not the same population as Reinold. The wiki leaves the two side by side. Lighting a muscle is not the same as changing it.
Stretch the joint you mean
For people who had already lost internal rotation, local posterior-shoulder work moved the joint in one session. Swanson, 2024, forty people: a grade-III posterior glenohumeral mobilization and a sleeper stretch both improved internal-rotation and horizontal-adduction range. The two local interventions did not beat each other.
Kang, 2019, another forty with an internal-rotation deficit: cross-body stretching alone helped. Adding a simultaneous dorsal-glide mobilization helped more — six degrees of internal rotation, ten degrees of horizontal adduction, and two centimeters on a mobility measure, versus stretch alone.
Pectoralis stretching shows up in two roles. Reiner, 2023, thirty-eight healthy adults, seven weeks, three times a week, fifteen minutes of pec-major static stretch: shoulder-extension range up about 6 percent, long-length isometric torque up about 11 percent. Stiffness did not change. Controls did not move. That is a healthy-adult mobility trial, not a desk-posture trial. The desk pairing is pec-minor stretch plus the muscle that pulls the blade back.
The cheaper half of the pair
Sepehri’s 2024 systematic review, twenty-two studies, reports that therapeutic exercise reduced forward-head, rounded-shoulder, and thoracic-kyphosis angles in people labeled with that upper-crossed cluster. The authors say strength, stretching, shoulder-based, and especially comprehensive programs “may be effective.” The abstract does not print the exercise lists.
Office-worker trials name the class more often than the sets. Depreli, 2024: sixty people with shoulder protraction; eight weeks of shoulder-stabilization work on top of office exercise beat office exercise alone on protraction, more strength sites, proprioception, and a closed-chain test. Go, 2016: thirty-eight office workers, six weeks, forty minutes, twice a week; scapular-stability work beat a manual-therapy comparator on forward-head distance, rounded-shoulder posture, upper-limb stability, and lower-trapezius thickness on ultrasound. Yaghoubitajani, 2022: both an online-supervised arm and a workplace corrective-exercise arm improved pain and photogrammetry angles from baseline; the supervised arm also separated from control on workability and upper-trap activation.
On rounded-shoulder posture specifically, stretch-alone lost the pec-length comparison. Lee, 2015, fifteen young men: scapular posterior-tilt after pec-minor stretching, and the same tilt with a brace, reduced the posture score and lengthened a pec-minor index more than tilt alone. Hasan, 2023, sixty young women: lower-trapezius strengthening plus pec-minor stretching increased pec-minor resting length more than stretching alone. Flexion range improved in both of Hasan’s arms; that one did not separate. Lee’s brace arm is a lab comparison, not a product review.
Pain and rotator-cuff-related reviews came back in the third search round and were set aside. Treatment of pain is out of scope. Clinical decisions stay open. A later pass can either find the named-lift bake-off the first abstracts did not return, or decide that kinds are enough. Until then the reader question is the same one the wiki left on the table: which kinds of work the papers support — not which workout to start on Monday.