Our attempts to understand and explain complex systems commonly involve reductionist analysis, whereby a system is broken down into its component parts for easier understanding. Theoretically, after analyzing and understanding each of the individual components of the system, we can “add” these understandings back together in a linear fashion and emerge with a complete understanding of the whole. For basic problems and many of our man-made systems, this is often a perfectly adequate approach.
But when it comes to more complex biological systems, like humans, purely reductionist approaches have led us to understandings that have ultimately proven incomplete and, in many cases, incorrect. This is particularly apparent in biomedicine where a number of diagnostic and therapeutic approaches that seem perfectly sensible under traditional reductionist thinking ultimately prove ineffective or even harmful (for more on this, including specific examples, see HERE as well as our podcasts and articles on pain).
In the context of sport and other physical endeavors including strength training, athletes and coaches apply reductionist thinking to the human body as a simple mechanical system in order to deduce conclusions about how humans should move for optimal performance and to reduce the risk of injury. This typically involves modeling humans as machines and applying basic physics and classical mechanics to a given task or athletic scenario. For example, the use of stick-figure free body diagrams are used to deduce conclusions about universal “optimal” movement patterns under load, as well as presumed risk factors for injury such as spinal flexion or knee valgus.
Furthermore, this modeling of humans as machines with “correct” and “incorrect” movements naturally leads us to conclude that skill-based tasks should be performed not only in accordance with the “optimal” mechanical model, but also be performed identically every time, i.e., with zero variation between trials. For example, an elite athlete should demonstrate technical execution that fits the predicted biomechanical ideal, and, due to their refined skills, should ideally execute it in this exact fashion on every attempt with no variation. This traditional view holds that any deviation from the ideal represents a movement error that should be corrected in order to improve performance; often, this reasoning is taken a step further to suggest that such deviations represent risk factors for pain and injury as well.
This is the view many assume to be true because “it just makes common sense”. However, as it turns out, the past several decades of research into human movement have brought this “common sense” into question.
- A high degree of inter-individual and intra-individual movement variability has been demonstrated across a variety of sports; in fact, it has been noted that “even elite athletes cannot reproduce identical movement patterns after many years of training, contradicting the ideas of motor invariance”. Bartlett 2007
- Movement variability appears to be beneficial in a number of ways (including, but not limited to, improving motor learning and mitigating injury risk); therefore, coaches should likely not impose excessive, perfectionistic constraints to eliminate movement variability in the early stages of learning, but rather allow individuals to explore potential solutions and to “self-organize” movement strategies over time.
- This study found significant inter-individual and intra-individual variability in the successful execution (barbell velocity profiles and kinematic assessment) of a 1-repetition maximum squat and six single repetitions at 90% of 1RM in reasonably well-trained lifters, which is in line with the evidence from a variety of other sports.
Nikolai Bernstein was a Soviet neurophysiologist with special interest in motor learning and motor coordination who, in the 1960s, published his text “The Coordination and Regulation of Movements”. He articulated the Degrees of Freedom problem, which states that there are numerous ways for humans to accomplish the same motor task due to an abundance of degrees of freedom at multiple levels in the neuromusculoskeletal system. This has led us to understand, as Bartlett et al. state that, “analysing discrete variables from isolated joints does not effectively capture the complexity of the coordinated motions of components of the body”. Research continues today into the question of how, specifically, we select a particular strategy among nearly infinite options in order to accomplish a given task.
For example, analysis of the biomechanical strategies of a variety of athletes, ranging from recreational-level to national and internationally competitive track & field athletes, basketball players, and many others has revealed that not only is there no single biomechanical ideal that reliably produces optimal performance across individuals performing the same task, but also that “even elite athletes cannot reproduce identical movement patterns after many years of training, contradicting the ideas of motor invariance”. Bartlett 2007
Following this line of research, modern theories such as Dynamical Systems Theory (DST) now challenge the traditional views of movement variability as “noise” or “error” that should be minimized. These more recent views hold that movement variability may actually be beneficial in a number of ways, and even vital for motor behavior. Latash 2012 For example, research from a variety of experimental models has shown that early movement variability actually improves motor learning efficiency by allowing for exploration of the wide variety of movement options. Wu 2014 Dhawale 2017
Motor variability is also necessary for changing coordination of movement in complex and variable environments, and likely offers benefits by distributing tissue loads more broadly across the tissues. The latter of these ideas has led to a “Variability-Overuse-Injury Hypothesis” based on a growing evidence base suggesting an association between decreased movement variability and injury, as well as evidence of a decrease in movement variability as task demands and loading increase. Nordin 2019
In this light, limiting movement variability may itself increase the risk of injury due to the cumulative effects of distributing load in the same way across the same tissues. Additionally, given that increasing loads and task demands necessarily tend to reduce movement variability (likely due to limiting movement options for successful task completion), this may explain one potential mechanism by which excessive exposure to high loading increases injury risk.
So, we have findings that suggest:
- Movement variability appears to be beneficial in a number of ways (including, but potentially not limited to, improving motor learning and mitigating injury risk)
- There is significant inter-individual variability in the execution of motor tasks (i.e., there is no single generalizable “optimal” motor strategy for all individuals to accomplish a given task)
- There is significant intra-individual variability in the execution of motor tasks, as evidenced by elite performers being unable to identically reproduce their own motor patterns across multiple trials of the same task.
It has thus been suggested that coaches should therefore not impose excessive constraints on individuals to reduce movement variability in the early stages of learning, but rather allowing them to explore potential solutions and ultimately to “self-organize” movement strategies over time. This has interesting implications when combined with our understanding of the Long-Term Athletic Development Model for youth athletes (which we discussed here, and is unfortunately outside the scope of this article). Even more recently, general research into weightlifting and powerlifting has increased, including research examining kinematics and movement variability of the competitive lifts that we’ll review from the article by Kristiansen et al.
The authors’ stated purpose was to “explore the level of inter- and intra-individual variability in the kinematic profiles of the back squat movement among skilled weightlifters”.
Study subjects included ten healthy males with at least 4 years of squat training experience, who could squat below parallel (defined as the crease of the hip below the top of the knee) and had no injury history or other compromising conditions.
- Age: 26 +/- 5.9 years
- Height: 1.83 +/- 0.08 m
- Body mass: 90.9 +/- 16.6 kg
- 1RM high bar squat: 165.5 +/- 27 kg
All subjects performed a standardized 1RM squat test protocol, followed by six single repetitions at 90% of 1RM, which were performed with 4 to 7 minutes of rest between repetitions. Subjects were required to begin from the upright position and squat until the crease of the hip dropped below the top of the knee (i.e., below parallel) in order to standardize the range of motion. Subjects used the high-bar squat technique. Any squats missing this depth requirement were not included in the dataset and the subjects repeated the attempt.
Subjects were studied using a Qualisys 3-dimensional motion analysis system, with kinematic tracking markers placed across a number of skeletal landmarks and on the barbell itself.
For analytical purposes, the squat movement was divided into five phases:
This is a caption
Correlation analysis was performed comparing each of the six repetitions performed at 90% against the 1RM in order to verify that the 90% repetitions were representative of the movement characteristics of the 1RM effort.
The following variables were calculated for each trial: vertical barbell velocity (VBarbell), angular velocity in the ankle (ωAnkle), knee (ωKnee), and hip joints (ωHip).
Finally, the following data were extracted from each of the six repetitions at 90% and averaged for each participant:
“Maximal anterior-posterior horizontal displacement of the barbell during the squat cycle, maximal range of motion of the hip, knee and ankle joint, mean vertical velocity during the AP phase, mean vertical acceleration and peak accelerations during the AP phase, Horizontal distance between barbell and hip joint, thigh length, and shin length.”
The authors performed multiple analyses on the data. First, they calculated Pearson correlation coefficients of barbell velocity profiles between each of the six trials at 90% 1RM for each of the 10 subjects. Next, they performed an analysis of variance (ANOVA) on these data (in short, this is a statistical method used to evaluate the differences between two or more samples), followed by a post-hoc test to determine where the specific differences occurred between groups.
The authors then calculated the mean barbell velocity profile for each of the participants’ six trials, along with standard deviations across all five phases of the lift. This allowed analysis of intra-individual variability at each phase of the lift. A similar ANOVA was performed across these data, followed by another post-hoc test to help “localize” any significant differences.
Finally, the authors performed regression analysis in an attempt to identify predictors of the intra-individual movement variability, comparing barbell velocity profiles against the anthropometric measurements and measurements of the individual movements listed above. In other words, are there anthropometric variables or lifting strategies (in terms of velocity profiles) that are more strongly correlated with (or predictive of) an individual’s movement variability when performing a heavy squat?
Table 1 below shows individual lifter mean velocity profiles and standard deviations, stratified by lifting phase. Overall, significant differences were found in vertical barbell velocity standard deviation between all lifting phases for all lifters. This is shown graphically in figure 1 below, averaged across all lifters with error bars depicting the substantial variability present.
Figure 1: Mean duration of each of the five lifting phases across six trials among ten subjects
Table 1: Data from subjects’ squats.
Due to the relative complexity of the data gathered, the data reflecting variability in angular velocity at the ankle, knee, and hip are presented exclusively in graphical form in the article (which we are unable to reproduce here) and would direct readers to Figure 2 in the original article for this evidence. In summary, the authors found significant inter-individual and intra-individual variability in these parameters as well.
There were three phases of the lift where little inter-individual variability in bar velocity was found; the end of the descent phase, the end of the sticking point phase, and the end of the deceleration phase at the top. This is explained by the commonalities among these phases where, under sufficiently heavy loads, bar velocity always approaches zero for all individuals. Interestingly, however, the kinematic data showed significant inter-individual variability in hip and knee movement at the sticking point, suggesting that lifters were utilizing different strategies for getting through this point in the lift.
Anthropometric analysis showed that shorter trunk and thigh segments, and/or longer shin segments, were associated with more intra-individual movement consistency (i.e., lower variability). Greater anterior-posterior displacement of the barbell was similarly associated with more intra-individual movement consistency.
What’s the Take Home Message?
This study found significant inter-individual and intra-individual variability in the successful execution (barbell velocity profiles and kinematic assessment) of a 1-repetition maximum squat and six single repetitions at 90% of 1RM. This finding should be unsurprising to anyone; however, the interpretation of such findings is where disagreement may arise.
The traditional view would hold that these variations, while unsurprising given that humans are not robots, represent “error” or “noise” that would ideally be minimized in order to improve performance. In contrast, the modern view (for example, as described by Dynamical Systems Theory), would argue that this variability is 1) desirable, 2) functional (i.e., serves useful purposes, as discussed in the introduction), and 3) likely cannot be eliminated, even among elite performers.
These data do not suggest that any one strategy is optimal (either across or within individuals). However, despite evidence from the Load Accomodation Model suggesting that movement variability reduces as load/task demands increase (likely due to fewer options for successful task execution), it seems that even under high relative loads there is substantial intra-individual variability in task execution.
This paper had several notable strengths and weaknesses; they used reasonably well-trained lifters who had an average 165 kg high bar squat, with stipulations that all squats be performed to below-parallel depth. This is by no means consistent with elite performance (especially at an average body mass of 90 kg), but represents a population that is likely far better trained (and thus, has had more skill practice with the movement) than the general population. The sample size of 10 is a limitation, obviously, although this was primarily intended as an exploratory study to lay the groundwork for future research. Furthermore, the repetition testing protocol involved performing repeat singles at 90% 1RM on 4 to 7 minutes of rest. For a very well-trained lifter, this can be quite challenging, although the authors note this and argue that 1) no correlation was observed between variability and trial number (i.e., variability did not consistently increase or decrease from trial 1 to trial 6), nor did bar speed significantly decline across the trials either. Finally, the statistical analysis involved normalization of each trial duration to a scale of 100 arbitrary time points; the authors recognize that this may have masked some additional variability in the timing of each phase execution.
Overall, these findings are consistent with much of the movement variability literature from other sporting contexts (including studies on international / elite athletes). It would be particularly interesting to have this sort of baseline data on an untrained population and to follow them prospectively across a long-term training program to periodically reassess their degree of movement variability, training outcomes, and injury risk, although such a study would be quite complex and costly to run.
To reiterate our arguments from the introduction; recent research suggests that movement variability is likely beneficial in a number of ways, and should not necessarily be viewed as a target for elimination. Rather, we should recognize what Latash calls the “Bliss of Motor Abundance”, that there are many redundant strategies available to successfully complete movement tasks, and that we should avoid excessive rigidity in our movement prescription for athletes — particularly during the motor learning stages, where we can allow individuals to “self-organize” their movement strategies and work collaboratively over time to optimize performance. We should also avoid demonizing movement variability from a pain or injury risk standpoint, as these have the potential to cause significant harm.
Kristiansen et al. Inter- and intra-individual variability in the kinematics of the back squat. Hum Mov Sci. 2019 Oct;67:102510.
Latash. “The bliss (not the problem) of motor abundance (not redundancy).” Experimental brain research vol. 217,1 (2012): 1-5.
Wu et al. Temporal structure of motor variability is dynamically regulated and predicts motor learning ability. Nat Neurosci 17, 312–321 (2014) doi:10.1038/nn.3616
Dhawale et al. “The Role of Variability in Motor Learning.” Annual review of neuroscience vol. 40 (2017): 479-498. doi:10.1146/annurev-neuro-072116-031548
Bartlett R. Is movement variability important for sports biomechanists? Sports Biomech. 2007 May;6(2):224-43.
Nordin AD. Reviewing the Variability-Overuse Injury Hypothesis: Does Movement Variability Relate to Landing Injuries? Res Q Exerc Sport. 2019 Jun;90(2):190-205.