Internal Rotation Peak Force Thoughts: What is the Lats Contribution?

Over the past year, my mentor and I have gone back and forth on the same problem more times than I can count: why is internal rotation (IR) peak force so hard to improve in certain overhead athletes?

It's not a niche concern. IR torque at the shoulder is one of the clearest markers of a high-performing shoulder – the benchmark sits around 2.1N/kg against just 1.26N/kg for the external rotators (Cools et al., 2016). That's not a small gap, and it's by design: overhead throwing demands far more from the internal rotators than the external rotators, not only to produce IR torque but also to control and resist anterior humeral glide.

So, we chased it. Isolated IR peak force work; however, the ForceFrame numbers didn't budge. Not for every athlete, but for enough individuals that it stopped looking like noise and started looking like a pattern. Rather than continually chasing the same adaptation, or, as Alex Wolf would say, 'applying familiar solutions', we started to ask a different question: what if the problem wasn't a subscapularis problem?

That sent us looking upstream, at moment arms, muscle architecture, electromyographic (EMG) timing and injury patterns to work out what the lat might actually be doing for the subscap during late-cocking and why the standard way we test and train the lat might be missing the point entirely.

What is the lat actually doing in the late-cocking phase?

During the late-cocking phase of a throw, as the humeral head migrates anteriorly-superiorly, the lats line of pull – via its insertion and attachment – imparts a posterior-inferior force to combat the initial anterior-superior glide that repositions the humeral head, altering the mechanical conditions the subscapularis is working under.

Six phases of throwing (Escamilla & Andrews, 2009)

This is not a torque story. Moment arm data makes that clear: the inferior subscapularis is the largest internal rotator of the shoulder by a distance, and the lat isn't a serious rotational competitor to it at all (Ackland & Pandy, 2011). What the lat does have is leverage in adduction and depression, and critically, the internal rotation moment arm it does possess is significantly larger when the humerus is adducted or flexed than when abducted and externally rotated (Ackland & Pandy, 2011), further confirming the lats' minimal contribution to IR torque when the arm is abducted and externally rotated during the cocking and acceleration phases of throwing compared to the pecs and subscap.

This is not only well established in the moment arm literature but also in throwing and striking EMG data, where we see the lat activity spike in the late-cocking and early-acceleration phase, not late-cocking and ball release (Escamilla & Andrews, 2009). One can therefore deduce that the lat's role is eccentric in nature during late-cocking: controlling the humeral head position to enable the subscap to express force under more favourable length-tension (L-T) relationships, rather than fighting a losing battle to prevent further anterior glide.

It is also worth highlighting that the lats and the upper (more phasic) portion of the subscap are peaking together in early acceleration. That's not proof of a casual synergy, but it's consistent with the idea that they're operating as a coordinated pair through the phase where humeral head control matters most, rather than as unrelated contributors.

Why does "tonic activity" need unpacking?

Without training the lat while trying to improve IR peak force, you're effectively trying to build the engine and anterior restraint to the shoulder without ensuring the humeral head is being positioned well enough within the glenoid in the first place. It would be like increasing the horsepower of a sports car without first checking the wheel alignment. The engine may be capable of producing more power, but poor alignment prevents that power from being transferred effectively into the road. Consequently, the subscap maintains excessive low-threshold tonic activity to ensure safety of the system, and that's what is potentially capping true peak force expression.

But "tonic activity" isn't inherently pathological – it needs separating into two things. Subscapularis is architecturally and neuromuscularly compartmentalised: the superior portion is fast-twitch dominant and functions as an early-onset phasic internal rotation, while the inferior portion is slow-twitch dominant and functions as a durable, tonic dynamic stabiliser resisting anterior humeral translation as its baseline job (Ackland et al., 2008). This is corroborated by cadaveric fibre-type mapping showing the inferior compartment running roughly 80% type I fibre against 78% type II in the superior compartment (Cho et al., 2023), with differential recruitment patterns between the two compartments further supported by Wickham et al.'s (2014) EMG data. That tonic drive in the inferior fibre isn't a fault state – it's design. Additionally, this dualistic function of the subscap is further reiterated by the muscle being innervated by two separate nerves (Sager et al., 2019).

However, what I am proposing here is that potentially excessive tonic drive, above that normal baseline level, is what happens when the humeral head is not being adequately positioned and controlled by its synergists. A poorly functioning lat may result in excessive anterior shear due to a lack of posteior-inferior force on the humerus, which keeps the subscap working overtime just to prevent further anterior shear and results in altered L-T relationships. Not only does this disrupt the force generation capabilities of the subscap, but it can also lead to a loss in range of motion (ROM) in the posterior cuff, which can be commonly seen in either cross-body horizontal adduction or internal rotation ROM evaulations.

Concentric Rings (Athletic Shoulder Instragram)

So – what inhibits subscapularis force expression? My working answer: the positioning of the humeral head. Unfavourable positions to produce force, driven by excessive tonic activity and an altered L-T relationship. As Dr Jordan Shallow mentioned in a recent webinar, 'position dictates function'; in other words, the position of the humeral head within the glenoid dictates the subscapularis' capacity to produce force, not the other way around.

Gary Ward's principle takes this a level deeper: 'Structure dictates function.' Before we even get to position, we have to ask what the bones and muscles actually built to do in the first place – and that's a question architecture answers more directly than anything else.

What does the muscle architecture tells us?

This is where the mechanistic story either holds up or falls apart, because architecture is one of the more direct predictors of what a muscle is actually built to do (Lieber & Fridén, 2000) – physiological cross-sectional area (PCSA) predicts force-generating capacity, and fascicle length predicts excursion potential (maximum distance a muscle can shorten and lengthen while producing force).

Lats: PCSA of 5.6cm², normalised fascicle length of 26.4cm (Gerling & Brown, 2013). That's a moderate force-generating capacity attached to a very large excursion range – the authors describe this explicitly as a muscle designed to produce a moderate amount of force over a large range of lengths. It isn't built to be a peak-torque producer at a single joint angle. It's built to stay mechanically active across a wide range of positions.

Subscapularis: PCSA of roughly 22cm², fascicle length of roughly 6.4cm with a pennation angle around 16° (Zhang et al., 2023, in vivo DTI architecture data). That's close to a fourfold larger PCSA than the lat, on a fraction of the fascicle length. This is a muscle built for high force output over a comparatively short excursion range – which fits its job as the dominant IR torque producer, not a muscle designed to track the humeral head through a large range the way the lat is.

Put those side by side, and the division of labour looks architecturally intentional: the lat is built to control and reposition across a wide range of humeral excursion, and the subscap is built to produce large, relatively short-range force once that positioning is sorted. That's a genuinely different line of evidence – independent of the EMG and moment-arm data – pointing at the same conclusion: the lat's job here is positional control and not torque-generating qualities.  

What does the injury data show us?

Understanding at which phase of a throw the lat is most commonly injured again may help further point us in the right direction to improving our understanding of when the most amount of stress is placed on the lats. We’ve already established from EMG data that the lat is most active during the late-cocking and early-acceleration phases of the throw (Escamilla & Andrews, 2009), where the lat is being loaded eccentrically, but this does not necessarily translate to higher loads placed on the muscle. 

If the lat were such a major torque producer in late acceleration and ball release, we’d probably not only see higher EMG activity, but injury data would no doubt point to higher injury occurrences in these phases, but it doesn’t. That is not to say the lat is not active during these other phases and doesn’t play a role; it’s just secondary (at least in my opinion!). To confirm this, a baseball study demonstrated that pitchers typically injure their lats during the late-cocking phase and early-acceleration phase of the throw (Erickson et al. 2019) – effectively during that elastic “whip” like motion where the arm rapidly transitions from layback into acceleration. It is at this stage where the anterior shoulder is placed under the greatest level of stress, 380N according to Fleisig et al. (1995), and consequently where the lat should work the hardest to control the humeral head. 

Late-cocking phase in baseball pitcher.

This injury distribution lines up precisely with the mechanical role I'm arguing for: not a phasic torque producer that gets hurt at peak concentric output (late-acceleration & ball release), but a control muscle that gets hurt exactly where it's being asked to eccentrically stabilise the humeral head under high contraction velocities at extreme ROMs.  

The testing & training problem!

So, firstly, how do we reliably decide whether the lat is contributing enough, in the way it needs to, in the positions it needs to, during overhead throwing actions? How do we test this quality? And most importantly, how do we train this physical quality?

Commonly, lat capacity is tested and trained with sagittal plane-biased humeral extension-based tasks such as rows and pull-ups. These are a reasonable convention, for reasons that have little to do with overhead activities. Humeral extension is the lat's largest, most dominant action in general daily and athletic tasks (think swimming, kayaking or rowing!); it's easy to standardise, and it has decades of normative strength data behind it. However, in my opinion none of that makes it correct for those specific questions above.

Moment arm data supports this scepticism. Ackland et al. (2008) found the lat's adductor moment arm in the frontal plane (peaking at -38.6mm in the middle sub-region) is substantially larger than its extensor moment arm in the sagittal plane (peaking at just -22.1mm, superior sub-region) – and this pattern holds up in Hik and Ackland's (2019) pooled review across studies (-37.3mm in frontal vs -26.8mm in flexion). In plain terms: the lat has more mechanical leverage to act as a humeral adductor than as an extensor. A bigger moment arm means more resistance and more tension development for the same external load – so a row loads the lat where its leverage and its training stimulus is comparatively smaller, missing the moment-arm conditions most relevant to late-cocking. That matters because it's precisely this tension-generating capacity, expressed in the frontal plane, that underpins the lat's ability to impart posterior-inferior force on the humerus during late-cocking.

Critically, as we have already established, the lat's role during late-cocking isn't concentric humeral extension at all – as the humerus is driven into horizontal abduction/extension and external rotation, the lat is being lengthened under load in the frontal plane, working eccentrically to help control humeral head position. This raises a mechanistic question – beyond direct evidence, but a reasonable hypothesis given moment arm and contraction-type differences: does a sagittal-plane row, with its concentric-extension bias, train a meaningfully different physical quality of the lat than the eccentric, frontal-plane adduction/depression demand placed on it during late-cocking? The SAID principle would predict yes, but this remains untested directly; no study has correlated frontal-plane eccentric lat capacity improved with greater IR torque output.

Concluding thoughts

The lat isn't a torque contributor competing with the subscap for IR output – the moment arm data rules that out. It's a positioning muscle, architecturally built for excursion over force, active through exactly the phases where the humeral head needs controlling, and injured in exactly the phases where that control is being demanded eccentrically at speed. If that's right, then testing and training need to reflect it.

Practically, this points toward something like a frontal-plane, 90/90-positioned isometric adduction test – the humerus at 90° abduction and 90° external rotation, loading the lat into adduction rather than the humerus in extension. That position isn't arbitrary: it sits close to where Ackland et al. (2008) found the lat's frontal-plane adductor moment arm actually peaks (around 70° abduction), and it approximates the joint position the shoulder occupies during late-cocking itself – exactly where we've argued the lat is doing its true work. Tested this way, in theory, we'd capture the lat's capacity to control the humeral head under load in the plane and position it's actually built to in, rather than measuring how much weight it can move through extension, a plane it rarely sees during the throw.

Rows and pull-ups aren't wrong; they still have a role, particularly early off-season to build general strength and CSA. They're just answering a different question than the one overhead athletes potentially need answered. That question being, can the lat control the humeral head eccentrically, and the plane and position – it's actually asked to do so during the throw?

That doesn't mean the test itself needs to resemble a throw – it doesn't, and probably shouldn't. A 90/90 isometric adduction test doesn't look at all like late-cocking of a throw; for starters, this hypothetical test is isometric! But it doesn't need to look like the skill to be meaningful; it needs to isolate and/or load the specific movement or physical qualities the skill actually demands. Specificity in testing isn't about mimicking the sport – it's about targeting the mechanism the sport relies on.

As with most things in this field, there's no universal answer. It depends on the individual, the demands of their sport, and a genuine read and an understanding of test data – not a proposed test replacing existing convention before it earns a place in the toolbox.

Because if the engine's fine and the numbers still won't move, maybe you've been checking and fixing the wrong part of the car.

References / Sources:

Ackland, D. C., & Pandy, M. G. (2011). Moment arms of the shoulder muscles during axial rotation. Journal of Orthopaedic Research, 29(5), 658-667.

Ackland, D. C., Pak, P., Richardson, M., & Pandy, M. G. (2008). Moment arms of the muscles crossing the anatomical shoulder. Journal of anatomy, 213(4), 383-390.

Ben Ashworth – Found of Athletic Shoulder

Cho, T. H., Hong, J. E., & Yang, H. M. (2023). Neuromuscular compartmentation of the subscapularis muscle and its clinical implication for botulinum neurotoxin injection. Scientific Reports, 13(1), 11167.

Cools, A. M., Vanderstukken, F., Vereecken, F., Duprez, M., Heyman, K., Goethals, N., & Johansson, F. (2016). Eccentric and isometric shoulder rotator cuff strength testing using a hand-held dynamometer: reference values for overhead athletes. Knee Surgery, Sports Traumatology, Arthroscopy, 24(12), 3838-3847.

Dr Jordan Shallow – Pre-Script Co-Founder

Erickson, B. J., Chalmers, P. N., D’Angelo, J., Ma, K., & Romeo, A. A. (2019). Performance and return to sport after latissimus dorsi and teres major tears among professional baseball pitchers. The American Journal of Sports Medicine, 47(5), 1090-1095.

Escamilla, R. F., & Andrews, J. R. (2009). Shoulder muscle recruitment patterns and related biomechanics during upper extremity sports. Sports medicine, 39(7), 569-590.

Fleisig, G. S., Andrews, J. R., Dillman, C. J., & Escamilla, R. F. (1995). Kinetics of baseball pitching with implications about injury mechanisms. The American journal of sports medicine, 23(2), 233-239.

Gary Ward – Founder of Anatomy in Motion

Gerling, M. E., & Brown, S. H. (2013). Architectural analysis and predicted functional capability of the human latissimus dorsi muscle. Journal of anatomy, 223(2), 112-122.

Hik, F., & Ackland, D. C. (2019). The moment arms of the muscles spanning the glenohumeral joint: a systematic review. Journal of anatomy, 234(1), 1-15.

Lieber, R. L., & Fridén, J. (2000). Functional and clinical significance of skeletal muscle architecture. Muscle & Nerve: Official Journal of the American Association of Electrodiagnostic Medicine, 23(11), 1647-1666.

Sager, B., Gates, S., Collett, G., Chhabra, A., & Khazzam, M. (2019). Innervation of the subscapularis: an anatomic study. JSES Open Access, 3(2), 65-69.

Wickham, J., Pizzari, T., Balster, S., Ganderton, C., & Watson, L. (2014). The variable roles of the upper and lower subscapularis during shoulder motion. Clinical Biomechanics, 29(8), 885-891.

Zhang, Y., Herbert, R. D., Bilston, L. E., & Bolsterlee, B. (2023). Three-dimensional architecture of the human subscapularis muscle in vivo. Journal of Biomechanics, 161, 111854.