Antagonist Co-Contraction: The Hidden Brake on Strength

Every time you extend your knee against real resistance, your hamstrings do not go quiet. They fire too, in the opposite direction, bracing the joint even as they oppose the very force you are trying to produce. This is antagonist co-contraction, and it is one of the least talked-about variables in how much weight you can actually move. It is not a flaw in your technique. It is a standing feature of how your nervous system manages every loaded joint you own.

The question that matters for anyone tracking strength progress is whether this braking pattern changes with training, and if so, in which direction. The honest answer, once you look past the textbook version of the story, is more complicated and more useful than most coaching content admits.

What Co-Contraction Actually Is

Every joint movement has an agonist, the muscle producing the intended force, and an antagonist, the muscle on the opposite side of the joint. During a leg extension, the quadriceps are the agonist and the hamstrings are the antagonist. Surface electromyography routinely shows the hamstrings are not silent during a maximal quad contraction. They activate at a lower level, simultaneously, and that activity opposes net knee extension torque.

This is not wasted motion. Co-contraction stiffens the joint, protects ligaments from shear forces, and fine-tunes movement precision, particularly at end ranges and under unfamiliar loads. The tradeoff is that any force the antagonist produces has to be subtracted from what the agonist can express externally. A more heavily co-contracted knee extension will always underperform its true quadriceps capacity on a force plate or a load cell, because part of what the quads produce is being cancelled out in real time.

The Classic Finding: Training Turns the Brake Down

The foundational study here is now more than three decades old. Carolan and Cafarelli trained the knee extensors of sedentary men with eight weeks of isometric maximal contractions and found extensor strength rose by nearly a third with no change in quadriceps EMG amplitude, a classic signature of a neural rather than a hypertrophic gain. The more specific finding was that hamstring co-activation during the extension effort dropped by roughly 20 percent, and the drop was measurable after the very first week of training.¹ A smaller version of the same effect even showed up in the untrained leg, consistent with a centrally mediated adaptation rather than a purely local, muscle-specific one.

A 2006 review synthesising decades of neural-adaptation research framed this as one of several ways the central nervous system can express more strength without adding tissue: alongside increased motor unit firing rate, doublet discharges, and reduced bilateral deficit, a reduction in antagonist co-activation was listed as a genuine, if inconsistently observed, contributor to early strength gains.² That review's central caveat has aged well: the nervous system is not simply trying to maximise force output at any cost. It is also trying to protect the joint, and those two goals can pull in opposite directions.

The Meta-Analyses Complicate the Story

Here is where the tidy narrative runs into trouble. When researchers pooled the co-activation literature specifically in older adults, the picture stopped looking like a reliable training effect. A 2014 systematic review and meta-analysis of seventeen studies found strength training reliably improved voluntary muscle activation of the agonist, but found no significant overall change in antagonist co-activation at either the ankle or the knee.³ A follow-up meta-analysis of randomised controlled trials in middle-aged and older adults reached the same conclusion in 2021: resistance training improved voluntary activation, but produced no measurable change in EMG amplitude or antagonist coactivation across 27 pooled studies.⁴

The reverse experiment tells a similarly nuanced story. If you deliberately train co-contraction itself, by performing repeated maximal simultaneous contractions of opposing muscle groups, does that raise your baseline co-activation during normal lifting? Two separate trials asked this directly and both found no. A four-week co-contraction training programme increased isometric elbow flexion and extension strength without any corresponding rise in antagonist co-activation during contractions against external resistance.⁵ A six-week version of the same protocol produced significant strength gains in both the flexors and extensors with no measurable increase in antagonist EMG, and no cost to rate of force development.⁶ Co-contraction, in other words, appears to be somewhat decoupled from strength training in either direction: general resistance training does not reliably wind it down, and specifically training it does not reliably wind it up.

Why the Relationship Depends on Who, and What, You Train

The picture sharpens once you stop averaging across populations and start looking at specific groups and specific exercises. A ten-week deadlift training trial in previously untrained adults found large effect sizes for reduced hamstring coactivation in both sexes, but the strength and activation gains were consistently larger in novice females than in novice males, suggesting starting neuromuscular status matters as much as the exercise itself.⁷ A twelve-month whole-body resistance programme in men over 70 raised plantarflexor strength and agonist activation substantially, yet antagonist coactivation at the ankle did not budge over an entire year of training, reinforcing that this adaptation is neither universal nor guaranteed simply by lifting for long enough.⁸

Mechanistically, the story is not just spinal. A single heavy-loaded training session produced measurable shifts in corticospinal excitability in both the trained agonist and its antagonist, but the deeper intracortical circuitry, the inhibitory and facilitatory networks that fine-tune movement, only adapted in the agonist. The antagonist's cortical machinery stayed unchanged after one session, hinting that agonist and antagonist muscles are not simply mirror images of the same adaptive process, they may be running on different timelines entirely.⁹

When Co-Contraction Is a Feature, Not a Bug

All of the above treats co-contraction as friction to be minimised, which is only half the picture. At the knee specifically, a healthy hamstring-to-quadriceps co-activation ratio is protective, and an imbalanced one is a recognised, modifiable risk factor for anterior cruciate ligament injury. A systematic review of exercises performed by healthy females found that bilateral squats generally produced poor hamstring-to-quadriceps ratios dominated by quad activation, while single-leg exercises performed through a moderate range of knee flexion produced healthier, more protective ratios.¹⁰ A gait study extended this further, finding that in females specifically, a weaker hamstring relative to quadriceps strength was directly associated with greater lateral co-activation during the loading phase of walking, a pattern with plausible implications for valgus loading and ligament strain, an association that did not appear in males at all.¹¹

The clinical picture reinforces the point from the opposite direction. A meta-analysis of 34 studies comparing people after ACL reconstruction to healthy controls found consistently elevated hamstring EMG amplitude and elevated hamstrings-to-quadriceps co-activation during gait and stair tasks in the reconstructed group, a compensatory pattern that persists well into rehabilitation and carries implications for reinjury risk.¹² Co-contraction, read this way, is not simply a brake to be released. It is a dial that needs to sit in the right range for the joint and task at hand, and where that range sits depends heavily on injury history, sex, and the specific movement being trained.

What This Means in Practice

A single surface EMG sensor placed over one muscle cannot see the antagonist directly, and no wearable should claim to measure a full coactivation ratio from one electrode site. What a wearable worn consistently on the same muscle across a training block can surface is something narrower and still useful: how much EMG amplitude a given external load is producing over weeks, at the same effort level, in the muscle you actually placed it on. If your quadriceps output at a fixed squat load is trending upward while the external load and rep quality stay stable, part of that story could plausibly include a more efficient agonist-antagonist relationship, alongside the usual candidates of improved motor unit recruitment and rate coding. The signal does not isolate the mechanism, and it should not be read as if it does. What it offers is a session-over-session comparison against your own baseline rather than a single-number verdict, which is the more defensible way to use any EMG-derived metric given how unsettled the underlying co-contraction literature still is.

For anyone selecting accessory work with knee health in mind, the exercise-selection evidence above is more actionable than the general co-contraction question. Single-leg work through a moderate range of motion produces more balanced hamstring-to-quadriceps activation than bilateral squatting alone, which is a reasonable input into programming decisions independent of any device.

Key Takeaways

• Antagonist co-contraction is your nervous system simultaneously bracing the opposing muscle during a lift, and it measurably subtracts from the net force your agonist can express externally.

• The classic 1992 finding that eight weeks of isometric training cut hamstring co-activation by roughly 20 percent, without any hypertrophy, remains a genuine and influential result, but it has not replicated as a general rule.

• Pooled meta-analyses in older adults found no significant overall change in antagonist coactivation with resistance training, and separate trials found deliberately training co-contraction does not raise it either, so this adaptation looks conditional rather than automatic.

• The effect appears to depend on sex, age, and exercise choice, with deadlift training showing larger coactivation reductions in novice females and cortical adaptations differing between agonist and antagonist after a single session.

• At the knee, co-contraction is not purely a brake to minimise, a healthy hamstring-to-quadriceps ratio protects against ACL injury, and single-leg exercises through a moderate range produce more protective ratios than bilateral squats.

References

1. Carolan, B., & Cafarelli, E. (1992). Adaptations in coactivation after isometric resistance training. Journal of Applied Physiology, 73(3), 911-917. https://doi.org/10.1152/jappl.1992.73.3.911

2. Gabriel, D. A., Kamen, G., & Frost, G. (2006). Neural adaptations to resistive exercise: Mechanisms and recommendations for training practices. Sports Medicine, 36(2), 133-149. https://doi.org/10.2165/00007256-200636020-00004

3. Arnold, P., & Bautmans, I. (2014). The influence of strength training on muscle activation in elderly persons: A systematic review and meta-analysis. Experimental Gerontology, 58, 58-68. https://doi.org/10.1016/j.exger.2014.07.012

4. James, E., Nichols, S., Goodall, S., Hicks, K. M., & O'Doherty, A. F. (2021). The influence of resistance training on neuromuscular function in middle-aged and older adults: A systematic review and meta-analysis of randomised controlled trials. Experimental Gerontology, 149, 111320. https://doi.org/10.1016/j.exger.2021.111320

5. Driss, T., Serrau, V., Behm, D. G., Lesne-Chabran, E., Le Pellec-Muller, A., & Vandewalle, H. (2014). Isometric training with maximal co-contraction instruction does not increase co-activation during exercises against external resistances. Journal of Sports Sciences, 32(1), 60-69. https://doi.org/10.1080/02640414.2013.805238

6. Zbidi, S., Zinoubi, B., Hammouda, O., Vandewalle, H., Serrau, V., & Driss, T. (2017). Co-contraction training, muscle explosive force and associated electromyography activity. Journal of Sports Medicine and Physical Fitness, 57(6), 725-733. https://doi.org/10.23736/S0022-4707.16.06363-5

7. Stock, M. S., & Thompson, B. J. (2014). Sex comparisons of strength and coactivation following ten weeks of deadlift training. Journal of Musculoskeletal & Neuronal Interactions, 14(3), 387-397.

8. Morse, C. I., Thom, J. M., Mian, O. S., Muirhead, A., Birch, K. M., & Narici, M. V. (2005). Muscle strength, volume and activation following 12-month resistance training in 70-year-old males. European Journal of Applied Physiology, 95(2-3), 197-204. https://doi.org/10.1007/s00421-005-1342-3

9. Mason, J., Howatson, G., Frazer, A. K., Pearce, A. J., Jaberzadeh, S., Avela, J., & Kidgell, D. J. (2019). Modulation of intracortical inhibition and excitation in agonist and antagonist muscles following acute strength training. European Journal of Applied Physiology, 119(10), 2185-2199. https://doi.org/10.1007/s00421-019-04203-9

10. Dedinsky, R., Baker, L., Imbus, S., Bowman, M., & Murray, L. (2017). Exercises that facilitate optimal hamstring and quadriceps co-activation to help decrease ACL injury risk in healthy females: A systematic review of the literature. International Journal of Sports Physical Therapy, 12(1), 3-15.

11. Smith, S., Rush, J., Glaviano, N. R., Murray, A., Bazett-Jones, D., Bouillon, L., Blackburn, T., & Norte, G. (2021). Sex influences the relationship between hamstrings-to-quadriceps strength imbalance and co-activation during walking gait. Gait & Posture, 88, 138-145. https://doi.org/10.1016/j.gaitpost.2021.05.019

12. Sherman, D. A., Glaviano, N. R., & Norte, G. E. (2021). Hamstrings neuromuscular function after anterior cruciate ligament reconstruction: A systematic review and meta-analysis. Sports Medicine, 51(8), 1751-1769. https://doi.org/10.1007/s40279-021-01433-w