Skip to main content

Why Do People’s Knees Cave Inward When They Squat?

Strength coaches and physical therapists tend to use fancy terminology to describe knee caving in a squat. For example, the terms knee valgus, valgus collapse, and medial knee displacement are tossed around quite frequently. Most strength coaches believe knee caving to be undesirable from a knee health standpoint. Countless greats in S&C circles seem to fall into this camp, including experts ranging from Kelly Starrett, to Louie Simmons, to Dan John, to Mike Boyle, to Mark Rippetoe, to Eric Cressey, to Tony Gentilcore, to Mike Robertson. It is thought that keeping the knees tracking over the toes in the squat will produce the least internal load on the passive knee structures, thereby keeping them healthy. Another champion of the knees out strategy is yours truly. I’ve written articles and filmed plenty of videos on this topic, including THIS one addressing valgus collapse as a whole, THIS one showing a simple squat correction, THIS one discussing coaching cues. Also, THIS guest blog from Derrick Blanton shows another simple correction strategy, and HERE is an article by my colleague Chris Beardsley discussing the mechanisms of knee valgus.

knees in-out

Strength expert and physical therapist Kelly Starrett showing knees-in versus knees-out in a squat


However, over the past couple of years, the “knees out” mantra has been challenged by some highly competent and knowledgeable lifters. For example, HERE is raw powerlifter Dan Green casting doubt on the efficacy of the knees out strategy, and HERE is the Chinese Olympic Weightlifting team’s take on knees out. These folks believe that allowing the knees to come in slightly at the bottom of the squat is safe and produces a stronger squat. A couple of years ago, I coined the term “valgus twitch” to describe the tendency of very strong lifters to utilize a quick “knees in” strategy when emerging out of the hole during squats. You see this all the time with powerlifters and weightlifters.

It should be pointed out that this debate does not apply to beginners. When beginners first learn the squat, the vast majority of them cave inward at the knees. After sufficient practice with sound technique, most lifters are able to regroove their motor patterns and their automatic default squatting technique will shift to a knees out strategy. I would also argue that women tend to cave inward even more so than men. In fact, every single female client I’m currently training battles the tendency to medially displace their knees when squatting. Knee caving is therefore a natural tendency of human movement. Beginners do it. Women do it. And even some of the most advanced male lifters do it.

However, there is a difference between beginners and advanced lifters regarding knee caving in that beginners will tend to cave inward via a combination of hip adduction, hip internal rotation, tibial external rotation, and foot pronation, and they’ll tend to exhibit this valgus collapse during a more broad portion of the movement. In contrast, more advanced lifters seem to briefly “twitch” into hip adduction when at the bottom of the squat as they initiate the concentric portion of the movement, and they usually revert back to having their knees out after they pass the sticking region.

Then Why Do We Cave?

Recently, Jonnie Candito of Candito Training HQ collaborated with my intern Andrew Vigotsky in order to tackle this question. Please watch the following video for a breakdown of possible factors as to why we might cave inward at the knee during squats. I found myself nodding in agreement throughout the entire video.

My Take – Why Do I Think Strong Lifters Cave?

Many beginners cave inward at the knees during a squat on account of insufficient ankle dorsiflexion range of motion or very poor gluteal strength and development. But we’re not talking about beginners right now. We’re talking about advanced lifters.

If an individual can squat with sound knee mechanics with light loads and keep the knees tracking over the toes, then this means that they do not have a mobility, stability, strength, or coordination problem. Their ankles are not too tight. Their hips are not too tight. The muscles that supinate the feet, invert the ankles, and abduct/externally rotate the hips are strong and stable enough. Some of my powerlifting friends exhibit knee valgus when they’re squatting. Am I supposed to inform them that they have weak glutes when they can squat, deadlift, and hip thrust over 500 lbs and lateral band walk like a boss? Methinks not.

Hi World Record Holder in the Snatch, did you know you had weak glutes?

Hi World Record Holder in the Snatch, are you aware that your glutes are weak? Yeah right.

There has to be some sort of strength advantage to caving inward at the knees, or else we wouldn’t resort to this strategy when maxing out or repping to failure. Since many lifters do in fact do this, we should try our best to understand it.

I have thought about this for many years and I’ve come to the conclusion that there has to be some sort of leverage improvement or lengthening associated with one or more of the hip extensors when the knees are caved inwards. In other words, when the hips are slightly adducted and internally rotated (valgus collapse), one or more of the muscles that extend the hips (the main muscles here are the gluteus maximus, hamstrings, and adductor magnus) would need to be stronger in this position. A muscle can be stronger if it is lengthened to a greater degree, more highly activated, or placed in a position with more favorable leverage (greater internal moment arm).

To my knowledge, there is currently no research investigating the hip extension torque/strength capacity of the various hip extensors at the bottom of the squatting movement in varying degrees of knee varus and valgus (different knee positions). Therefore, I’m unaware as to whether the gluteus maximus, hamstrings, adductor magnus, or hip external rotators are stronger as hip extensors with the knees exhibiting valgus collapse.

Side Note for Sports Science Students: In order to tackle this question, it would require an entire PhD thesis (hint hint aspiring PhD sports science or muscle modelling students). The first study could investigate 3D joint angle movements, EMG, sticking regions, and torque angle curves in the deep squat. The second could investigate hip extension torque in an isokinetic dynamometer with varying degrees of knee varus/valgus and also vertical force production on a force plate during an isometric deep squat with varying degrees of knee varus/valgus. The third could utilize musculoskeletal modelling to estimate moment arms and lengths of all main hip extensors throughout the hip flexion-extension movement continuum in varying degrees of knee varus/valgus. The fourth could be a training study to see if hip strengthening and motor control training improves upon squatting technique with maximum loads and helps eliminate knee valgus. But I digress…

I surmise that the lateral hamstrings (biceps femoris long head) would be stretched to a greater degree in knee valgus, whereas the adductors and possibly the medial hamstrings (semitendinosus and semimembranosus) but not the gluteus maximus or lateral hamstrings (biceps femoris long head) might have greater hip extension moment arms in knee valgus, but this would require research for verificaiton. A recent article published in Clinical Biomechanics with the title, “Lengths of the external hip rotators in mobilized cadavers indicate the quadriceps coxa as a primary abductor and extensor of the flexed hip” indicates that the hip external rotators (namely the piriformis and obturator internus) are in fact strong extensors of the hip, and their length is increased in hip flexion and adduction (meaning that they produce greater passive elastic force at the bottom of a squat, and this is enhanced when the knees are caved inward). Considering that the hips are more heavily relied upon as loads increase in the squat, it makes sense that lifters would resort to mechanical strategies that would enhance the strength of the hips with maximal squat loads.

My current position on the knees out debate is that:

  1. Beginners must learn the rules before they can break the rules and master the knees out technique when squatting
  2. A “valgus twitch” is acceptable for powerlifters and Olympic weightlifters as long as it’s not associated with knee pain
  3. Since many individuals appear to be stronger in slight knee valgus, this indicates that their hip extension torque production is stronger in slight hip adduction and internal rotation
  4. Future research should investigate the hip extension torque capabilities and partition the individual contributors to hip extension torque in deep hip flexion with varying levels of knee varus/valgus


  • Shannon Khoury says:

    Bret, this is one of the most interesting things I have read in a while. Thank you for sharing!

  • JP Rose says:

    My statement is purely conjecture, I believe when a valgus collapse is present it is a compensatory movement. I agree that people who lift a lot of weight are not weak. However, I would state that when you are going to max or close to that the valgus collapse is letting the lifter know that for that given load the glute medius etc….. may not be strong enough to handle it. I’m sure at loads that are 20 plus % lighter than max this probably doesn’t exist. I would try maybe hip airplane, strict form SLDL with heavy kettlebell, some RNT WORK WITH bands, heavy suit case carries etc…. and assess where things are. I would venture to guess most lifter will need a lot of improvement in those exercises. Because the are so strong in the Sagital plane they get away with it at lighter loads. I personally feel lifting any weight without near perfect form is a recipe for disaster that’s why owning the weight and movement pattern is a must before progressing forward.

    • Bret says:

      JP – just one point of contention…in deep hip flexion, the glute medius becomes a hip internal rotator, not a hip external rotator. Maybe midway the glute med is effective in preventing valgus collapse, especially some of the fibers. See here: I believe I elaborated on this article in one of the blogpost links I listed in the first paragraph of this article.

  • Tyler says:

    Brett I’ve also been suspicious that certain types of bad form offer great mechanical advantage. Things like rounding your back fling your elbows etc. come to mind. based on the area of the range of motion where this tends to occur I would assume the advantage is being offered to the quads. Although I’m unsure of the advantages is mechanical or neurological in nature.. During a extremely high effort squat it seems I only have enough neural drive to either push into the floor or to hold my knees out and push into floor. The knees in position seems to offer good enough passive stability that allows all of my muscular force to be directed towards my quads. If you try this now by pushing an inch or two out of the hole with your knees out you feel that both your quads and glutes light up whereas if the knees come in and you push into the ground only your quads seem to light up and it seems to a greater degree then before. This may be why we see this flinch during the first couple inches (where the quads are needed most). I’m sure there are elements that make the hip stronger too but it may require more specific inquiry to discover.

    • Bret says:

      Tyler, I was going to mention this. It is foolish of me to assume that the advantage has to be at the hips when it could also be present in the quads. I just can’t seem to wrap my brain around how the quads would get an advantage – I don’t think the rec fem would have better leverage, nor would the vastis, nor would any of them be stretched further. but this still could be a possibility so good call.

      • Tyler says:

        Just curious, but if I’m already finished with my decent and cave during the transition upward, how would a muscle being stretched to a greater degree (like the glute max, deep rotators, biceps fem in your example) aid in a concentric? It seems like hip extension is almost impossible in internal rotation while comin out of deep flexion. The sensation of caving seems to “shut my hips off ” instead of giving them more power. Just conjecture though.

  • Tyler says:

    Flaring*** not fling

  • Bryce says:

    Great stuff Bret! Seems to make sense that the safest position is the one in which the internal forces are overcoming the exertal forces: better to have the knees bow slightly that to rupture a quad or hamstring?

    So rather than cuing “knees out,” perhaps it’s best to just keep new lifters focused on glute training when they manifest leg bowing? I suppose what also must be taken into account is that this discussion applies primarily to athletes who choose to employ the barbell squat in the training. It seems very possible to have a successful training career without cultivating the back squat (heresy!).

    • Bret says:

      Sure Bryce, athletes definitely come to mind here as the primary goal is to keep athletes healthy, and they don’t get paid more to lift heavier weights. Their form should definitely be more strict than that of a powerlifter, especially when the powerlifter is on the platform and can allow for more wiggle room.

  • Always enjoying your work Bret.
    Always informative.

    Two of the key things Andrew mentioned regarding knee valgus were that it’s ok as long as it’s controlled and not going to the degree of stressing passive structures. How are these two determined by the Coach watching a person perform a lift? I do not allow clients (predominantly general population and some Sprinters) to perform a squat or lunge with the knees consistently going medial of the 1st MTP.

    I look forward to your reply.

    • Bret says:

      Odain, this is an incredible question, and I’m afraid I don’t have a good answer for you.

      How do we know how much buttwink is acceptable, how much spinal rounding is okay, how much knee valgus is fine, etc.? This is where the trained eye comes into play, and it takes many hours of coaching to obtain this. I believe that I’m very good at intuitively knowing what’s acceptable and what’s not, but we shouldn’t be arrogant since people have different genetics pertaining to soft tissue strength and some can’t get away with much.

  • strini says:

    Hi Brett
    As always it is very enlightening reading your articles.

  • Eric says:

    As always excellent excellent excellent. Is it possible elite lifters have this “valgus wink” to access their base of support better and get their knees under their center of mass while keeping their feet wide to maintain balance? I’d imagine that the length tension changes at the hip also play a role – but maybe these lifters are better able to utilize their passive structure (skeleton) so that their movers can do their job more efficiently. Just a guess. You’re the man!

    • Bret says:

      Hmmm, maybe this is a possibility Eric. But then this would imply that a shoulder width stance would be best for hoisting heavier loads, and most lifters are stronger squatting with slightly wider stances. But it could definitely be that the passive structures in the knee and hip stabilize the knee, leaving more room for the muscles to focus on raising the weight. Let the passive structures stabilize in the frontal/transverse planes so the prime movers can focus on saggital responsibilities.

  • Jeremy Rucker says:

    So there is a minimal to no account here of the variations of structure with different people’s bodies. The ‘Foldability’ in a squat depends on soooo many things: limb length proportions (trunk to femur to lower leg/foot), muscular inhibition in any area of the body that is decelerating you in a squat can allow for a number of compensations to perform the specific squat you set up, limb position (foot placement, structure of the shoe you are wearing), the application of the load matters.
    To also suggest the bicep femoris being excessively ‘stretched’ and the Glute Max not having a good moment arm comparatively to anything else that hip extends you from the bottom of the squat is also a haphazard assumption. Where is the load line in the squat? Is it more over the trunk with a larger moment arm to the knee, ankle? Is it more over the knee, with a lager moment arm to the trunk/spine and hip? Until that is determined, among determining where weaknesses come from, it is NOT possible to discern which muscles have larger moment arms to bring you up out of that squat.

    We ask complicated questions, expect simple answers to make ourselves sound smart. It might have been a better appropriately go through all the factors of a knee vagus stress in a squat, pro/con or reasons why. Even a Phd, would have a hard time coming up with an answer. There are infinite possibilities to consider for their reasons. Covering those possibilities would be a much better start.

    • Bret says:

      Jeremy – “foldability” as in sagittal plane? I’ve written many times about this (I even liken the torso/femur/tibia to a lightning bolt: But we’re not talking about sagittal plane stuff here – we’re talking about frontal/transverse plane stuff. I agree that structure plays a large role in determining the tendency to cave inwards at the knees, such as pelvic width/Q-angle, etc., which is also why women may have a greater tendency to valgus collapse, but since many people exhibit this tendency when maxing out, I didn’t address it in the sake of brevity.

      And by the way, I didn’t say that the biceps femoris is “excessively” stretched, just that it’s stretched more with internal rotation at the hips. And I didn’t say that the glute max didn’t have a good moment arm, just that it wouldn’t be improved with internal rotation (but I could be wrong – it would be stretched more but not sure if this would be meaningful).

      In addition, you’re talking about external moment arms (hip extension moment arms and knee extension moment arms) in a squat. I’m talking about internal moment arms (aka muscle moment arms) – see here:

      If you watch the video, my intern Andrew did go through all of the factors. That’s why I said, “here’s my take” and listed my thoughts. I don’t think there are an infinite number of possibilities, but it could definitely be due to multiple reasons (we tend to be reductionists and ignore the interrelationships of varying factors). So I agree with you about that.

  • Matthew Barba says:

    I’ve been a powerlifter for the majority of my life and I have had countless people, physical therapist, and know it all’s tell me that I have week abductors and my knees will give out on me and I will be lucky to ever squad again. But finally someone out there took the time to research and understand this syndrome. I applaud you for answering a very long question that I have been searching for thank you so much.


  • Neil Parsley says:

    I believe the same mechanical advantage (knee valgus) is utilised by elite sprinters when driving out of the blocks

  • Eric says:

    Great summary. As a coach who works predominantly to with athletes it’s important for me not to groove valgus movement patterns as acceptable in strength movements. Although valgus happens in sport, grooving a pattern where it’s acceptable and common opens a pandoras box of potential risks that outweigh the possible benefits for most athletes.

  • chris says:

    Great discussion guys – in addition to the advantages you discussed regarding hip extensor improvements in internal rotation, my guess is that additionally, at the bottom of the squat, the rectus femoris is slackened at the hip and can’t provide much force into knee extension – moving into a more valgus position wraps the patella / extensor mechanism around the lateral femoral condyle and allows the lateral and intermediate quad to utilize that lateral femoral condyle as larger lever into knee extension; until the athlete is out of deep knee flexion and less knee extensor moment is necessary to complete the task. That would have implications for the patella – compression over the lateral patellar facet, and patellofemoral pain, but these athletes have conditioned their patellar bone density over time to withstand that type of compressive force, and thus they can handle it without injury or pain.
    Thanks for posting!

    • christian says:

      Relations to your rectus femoris analogy, whilst the hip is relatively flexed, the knee is at maximal flexion (ass to grass). Would the rectus femoris still be on slack as you have suggested?

  • Corey Wesley says:

    what was the study mentioned in the video related to heel elevation and squatting? The year was 2007.

  • joyce says:

    Interesting article. Bret, any comments on why a moderately well trained, somewhat advanced, trainee would somewhat suddenly develop a caving problem in the deadlift? Would you think of it the same way as in the squat? (yes I have this problem!)

  • Hey Bret! With reference to athletes…elite athletes going into degrees of knee collapse during squat or relative deep squat end range simply be because of a quantitative compensation dependent on load…specifically heavy and near max loads. Load a body with increasing load at some point creates a threshold along and across all joints and their systems where millimeter by millimeter and of ALL joints involved from top to bottom ‘give up’ some relative and quantitative stabilization characteristics that eventually yields an end range gross response i.e. valgus?

    • …Maybe tissue deformation relative to high loads is what I was trying to suggest here…which would certainly drive a neural response or ‘twitch’…
      I appreciate your work Bret. I am sure you are not far from being recognized at #1 contender in our industry brother

  • christian says:

    Perhaps it’s a combination of bony alignment, neuromuscular fatigue under load and motor control issues.

    Example an individual with excessive external tibial torsion and or anteverted hips will find it much harder to control valgus migration inheritantly based on their bony configurations. Perhaps they are more likely to fatigue and less likley to tolerate excessively high loads.

    I notice those questioning the idea of reducing valgue stress are predominnatly olympic and powerlifters who lift appreciable amounts of weight. They’ll be the first to excuse slight lumbar flexion or hyperkyphotic thoracic position in maximal and submaximal deads so why wouldn’t it be any different here?

    I think it’s all relative to the individual ie their bony and soft tissue restrictions, training experience, sex (specifically acknowldeging Q angle variances), and load they are aiming to endure.

    A similar debate has existed for many years regarding sub talar joint pronation. You have old school thinkers believing sub talar joint neutral is the optimal position of function as opposed to those considering it a transitional position the rearfoot passes through during movement with the emphasis being placed on torque encountered during foot function as the bigger issue.

    I personally believe it’s likely both. You determine an individuals neutral first (dependant on bony development), understand how that then influence force distribution and then consider that in their movement assessment????

  • Rob Panariello says:

    My two cents… I agree with Matt Jennings statement. Often times with regard to finding a solution to various situations we are so focused to look for the “gold”, the “answer” so to speak, that as we are “mining” for this answer unnoticed are all of the other gems we pass along the way. The structural anatomy of an individual does not change during the performance of an exercise. Generally what changes over time (as the athlete improves) is the exercise prescribed volume and applied weight intensity, as well as enhanced muscle group contribution, joint/soft tissue stresses and perhaps levels of fatigue.

    If the athlete demonstrates appropriate squat exercise technique utilizing body weight and lighter loads then the anatomy allows for proper squat technique. As heavy loads are applied and resulting increased joint stresses as well as the enhanced contribution of different muscles groups which are emphasized at different points in the exercise range of motion, especially at the squat “sticking point” occur, squat form may deviate, as the body compensates for the unaccustomed heavy load. A loss of exercise range of motion (loss of squat depth) may occur as well.

    Anatomy varies in every individual yet many demonstrate proper squat form. Sometimes we have to look in the mirror and realize that we just weren’t strong enough to maintain proper squat exercise technique as specific heavy loads are applied.

    Just my opinion
    Rob Panariello

  • P Stevenson says:

    Buttwink? What is that? Posterior tilt? Pelvic floor contraction? That’s a new one for me!

  • Jim says:

    “Since many individuals appear to be stronger in slight knee valgus, this indicates that their hip extension torque production is stronger in slight hip adduction and internal rotation
    Future research should investigate the hip extension torque capabilities and partition the individual contributors to hip extension torque in deep hip flexion with varying levels of knee varus/valgus”

    I know this may be a blasphemous stance but if you know that the hip extensors produce maximal torque with the knees approximately hip distance apart shouldn’t that be where the feet are too to insure that hip extensor torque is delivered perpendicular relative to the floor. If the lower leg is anything other than perpendicular to the floor hip extensor torque is partially wasted. This is basic physics. Anything other than perpendicular and the force is expressed tangentially. Think about if you stepped off a high platform and landed in a squat postion with your feet shoulder width apart or greater. Sounds like a recipe for disaster as the hip extensor torque and impact would be absorbed to some degree by the connective tissue of the inner knee. When someone does land efficiently from a height the feet are in line with the knees and hips. The feet are under the hips because human being create and absorb force most effectively with feet under the hips.
    Same thing with a vertical jump. People don’t jump high with feet shoulder width apart. They jump high with their feet under their hips 6 to 8 inches apart.

    So why the discrepancy with the squat. I belive it is for two reasons 1) the emphasis on moving loads instead of properly training movement patterns. And 2) the difficulty of addressing the lack of range of motion needed to perform a squat properly, we thus allow for compensations like moving the feet wider to achieve the level of depth we persume is necessary for the squat to be legitimate.

    Lastly look and toddlers when they squat and stand. When standing their feet are under their hips, when squatting their feet are under their hips. They have the ability to comfortably eccentricly elongate their hip flexors and planter flexors to allow a full squat in an ideal positon. Since we know muscle length is more of a neurological function than a structual one reprogramming the neurology (completion of the alpha gamma loop) will yield an efficient squat. Depth of a squat can then be determined by the ability to eccentrically elongate the hip extensors and planter flexors while maintaining proper position.
    Just a thought

  • Tyson Schoolman says:

    i found the article very thought provoking. It might be possible that the answer lies in the synergistic movements involved in the diagonal flexion/extension pnf patterns. I looked at it in a neuro textbook and thought about for a while and found some similarities but couldn’t visualize it exactly in my head.

  • Yeah, very good!

    I’ve noticed this too in high-level oly and PLers. Makes sense that it would advantage them.

  • Ben says:

    Hi Bret — amazing work as always. Please continue with research, EMG studies, etc.

    My question is…what is your take on ONE knee caving in only? And what if it’s associated with something like a hip shift coming out of the hole (one knee can’t stay out while hips shift to that same side)? There’s also a tight or compressed QL present at the bottom of the squat? Thanks!

  • Charles Nankin says:

    Only my right knee strays in. Interestingly, my right leg is 20mm shorter than the left. Also my right hip had Perthy’s Disease when I was a boy – which is delayed development of the hip ball – and which finished when I was 16.

    Started squatting again 18mnths ago after 7yr hiatus and lumbar and other surgeries. Will try to do more grooving, adductor work, and squatting/thrusting with a band around the knees. My glutes and thighs are well developed. I do a full squat onto my ankles of about 250lbs for 5RM. 6’4″ and 200lbs at 37yrs.

  • Aaron Doyle says:

    Hi Brett,
    Always great to read your posts and listen to the podcast! Would it be at all possible that the Valgus squatting debate could be just a human reaction to accepting force production? Like when a footballer cuts, a weightlifter pronates, adducts and Internally rotates (shifts into a valgus state) to store elastic energy (stretch shortening cycle) into the hip External Rotators to help punch out of the bottom of the squat?

  • James See says:

    Great article,

    I have always been stuck in the middle between knees in or out debate, as I usually see athlete’s knees caving in on the heavy squats as they try to recruit more quads, but when i try it at the bottom of the squat, realized that my knees are hurting and now I know that knees valgus is not for everyone with different body types and mobility. Your advice came timely before i bust my knees, will stick to neutral knees and not cave in.

  • Goi says:

    How about when the knees cave only on 1 side, and only when the weight is high, or when fatigued? I have no problems at lower weights or when I’m not fatigued, but my left knee caves in and my hip shifts to the right at the end of long sets, presumably because of fatigue. What should I be strengthening and/or stretching then?

  • Brian says:

    I like the knees out approach to provide a que to open the hips. As a result I use a que to close my hips as I go back up. I feel it get me up stronger and faster. However When i go heavy, I feel as though I can break form and cave inward even more. So I have started to just follow a guideline of hips and where the knee is in relation to the ankle.

  • tseki says:

    In the Candito Video, Alex mentions that if your muscles on the lateral side of your legs are stronger, then an athlete might experience knee valgus. What can an individual do to strengthen the muscles on the medial side?

  • Kane says:

    Nice post! So this applies to advanced lifters, but what about novice lifters? I’ve just started squatting and I’ve found that my knees have a tendency (not all the time but some of the time) to cave in during the squat – this seems to occur equally on both the eccentric and concentric portion of the movement. Will this rectify itself naturally or are there things I should be doing/focusing on? Thanks in advance.

Leave a Reply


and receive my FREE Lower Body Progressions eBook!

You have Successfully Subscribed!