Nonlinear Analysis of an Unstable Bench Press Bar Path and Muscle Activation : The Journal of Strength & Conditioning Research

Secondary Logo

Journal Logo

Original Research

Nonlinear Analysis of an Unstable Bench Press Bar Path and Muscle Activation

Lawrence, Michael A.1; Leib, Daniel J.2; Ostrowski, Stephanie J.3; Carlson, Lara A.1,4

Author Information
Journal of Strength and Conditioning Research 31(5):p 1206-1211, May 2017. | DOI: 10.1519/JSC.0000000000001610
  • Free



Recently, training with an unstable load has become more popular in both the general fitness world and when training for specific strength sports. The assumption is that training with unstable loads will result in increased recruitment and adaptation of the stabilizer muscles, which is beneficial for sport performance and activities of daily life (7,9). To this extent, special flexible barbells have been created in an effort to increase the instability of a load. However, only 1 study has investigated the effects of these flexible barbells on muscle activation (11). Although there were some promising findings of increased stabilizer muscle activation, whether the load is truly made unstable, and the effect of this type of training on motor control has yet to be examined.

Although the research is limited, there have been some promising findings of increased muscle activation in both the upper extremities (11) and torso musculature (8) when using an unstable load. Parallel back squatting with elastic bands suspending the load has been shown to increase stabilizer muscle activity (rectus abdominus, external obliques, and soleus) (8). However, Dunnick et al. (3) found there to be no difference in muscle activation while bench pressing with the weight suspended from elastic bands versus bench pressing with a normally loaded barbell. Conversely, Ostrowski et al. (11) found that while benching with a load completely suspended by elastic bands onto a flexible barbell increased muscle activity of the biceps and the left middle deltoid . Interestingly, these increases in muscle activation occurred, although Ostrowski et al.'s (11) unstable load was 60% of 1 repetition maximum (RM), whereas the stable load was 75% of 1RM. These differences in findings may be due to different experimental setups; Dunnick et al. (3) imparted several constraints that may have decreased the effects of the unstable load on the participants (e.g., a slow controlled cadence, requiring a pause at the bottom of the press, only suspending part of the load with elastic bands, and using a standard barbell). Ostrowski et al. (11) had participants press a flexible barbell as fast and as safely possible with the entirety of the load suspended by elastic bands.

Previous studies examining muscle activation and unstable load exercises may have failed to indicate desired responses because the measures used may not have been sensitive enough to gain an insight into the full spectrum of responses. Increasing or decreasing muscle activation, while useful information to coaches and athletes, describes only a small aspect of the response unstable loading may elicit. Measurements of control should also be used to indicate how the lifter is responding to the challenge, as, while the task may or may not increase or decrease total muscle activity, the organization of that activity may also change. These organizational changes are relevant as they give insight into the pattern of muscle activation, not just the magnitude over time (2,4,12). To this end, nonlinear measurements of the dynamic stability of a system and the predictability of motion were used to evaluate potential changes in muscle activation control and organization.

Although unstable load training with a flexible barbell is already a very popular training method, and several companies currently manufacture flexible barbells, there have been no investigations concerning if these barbells actually create an unstable load or how they alter muscle activation organization. Therefore, the purpose of this study was to examine the effects of an unstable load (as provided by a flexible barbell and a load suspended by elastic bands) on the bar path, and the primary and stabilizing musculature while bench pressing using nonlinear analyses. Given that no other study has examined bar motion and muscle activity using a nonlinear approach, we hypothesized that (a) the unstable condition would produce greater bar path variability in all directions; (b) the stable condition would produce a more predictable bar path in all directions; and (c) during the unstable condition, muscles would activate in a more constrained pattern.


Experimental Approach to the Problem

The data presented in this article are part of a larger study collected for the purpose of investigating responses of muscle activation to unstable loading during the bench press. The purpose of this analysis was to gain a better understanding of the stability of unstable load bench pressing, and whether differences in muscle activity could be found where none were identified in a previous investigation (11) using a different analysis. Fifteen resistance-trained volunteers bench pressed 60% of their 1RM bench press under the unstable condition and 75% of their 1RM under the stable condition. Because of the increase in difficulty in the unstable condition, 1RM percentages were standardized. The 60% of 1RM is comparable with loads tested in previous studies where subjects squatted with unstable loads (8). The 75% load was used with the standard barbell as this is a typical load used for 5 repetitions. Bar position and muscle activity of both upper limbs and trunk were recorded and analyzed with nonlinear measures to determine bar path stability and predictability and if muscle activation patterns were constrained differently between stable and unstable conditions.


Fifteen resistance-trained men (age 24.2 ± 2.7 years, mass 84.1 ± 12.0 kg, height 1.77 ± 0.05 m, 9.9 ± 3.4 years of lifting experience, and bench press 1RM 107.5 ± 25.9 kg) volunteered for this study. Individuals with current upper extremity injuries, injuries that prevented them from exercise in the past 6 months, or those who have had upper extremity surgeries were excluded from participation. Seven subjects reported using unstable load training before participation in this investigation; however, their reported use was sporadic and unsystemized. This study was approved by the institutional review board (IRB), and all participants gave written informed consent (IRB #042815-014).


The testing methodologies used in this study are the same as reported by Ostrowski et al. (11). In short, all subjects were asked to complete 2 testing sessions. To minimize the influence of fatigue, subjects were asked to abstain from exercise for 48 hours before testing. During the first testing session, subjects performed a 1RM bench press with a standard barbell and typical load according to the National Strength and Conditioning Association's guidelines for maximal strength testing (1). The second testing session occurred at least 7 days after the 1RM bench press test and included the stable (75% 1RM) and unstable (60% 1RM) pressing conditions. The order in which the conditions were performed was randomized.

During the second testing session, bar path was tracked by measuring the position of a reflective marker on the right side of the bar, and muscle activation of the pectoralis major, anterior deltoid, middle deltoid, posterior deltoid, biceps, triceps, upper trapezius, and latissimus dorsi were measured bilaterally during 5 repetitions of both conditions. For the stable condition, weights were placed on the barbell normally. In the unstable condition, weights were suspended from “mini” elastic resistance bands (EliteFTS, London). The bands were “quadruple looped” through the weights and hung on the flexible Earthquake bar (Bandbell, Columbus, OH) (Figure 1). As a precautionary measure, the load on each band was limited to 50 pounds. If more than 50 pounds was needed to reach the prescribed load, then additional bands were used until the correct load was achieved.

Figure 1.:
Load suspended by elastic bands.

Nonlinear analysis methods were used to explore the way in which participants controlled how they lifted the bar in both conditions. To analyze electromyographic (EMG) data, a method similar to that used by Kang and Dingwell (5) was used to calculate local stability. In short, the time delay and embedding dimensions were determined using an automated algorithm based on the method described by Kennel et al. (6) and an average mutual information algorithm. The method of Wolf et al. (13) (Lyapunov exponent [LyE]) was then used to calculate the local stability of the signals. To describe control of the person-weight system kinematics, we used the LyE to calculate local stability of a bar marker trajectory and sample entropy to describe the self-similarity of the bar path. In pilot data exploration, both markers on the bar behaved very similarly, as may be expected, so only the marker on the right side of the bar was used in the final analysis. All nonlinear analyses were performed using custom MATLAB 2016a (Mathworks, Natick, MA) programs.

Statistical Analyses

Data from bar path LyE, bar path sample entropy, and muscle activation LyE between conditions were compared using a repeated-measures analysis of variance, resulting in 3 statistical tests. Statistical significance was set at the 2-tailed p ≤ 0.05 level of confidence. Statistical analyses were performed using SPSS IBM (SPSS Statistics 21; IBM Corporation, Somers, NY) software package.


The unstable condition produced significantly greater LyE values in the anteroposterior (p = 0.001) and mediolateral (p = 0.002) directions, but there was no difference between conditions in the Z direction (Table 1). For sample entropy, the unstable condition produced significantly smaller values in the anteroposterior (p < 0.001), mediolateral (p < 0.001), and Z directions (p = 0.001) (Table 2). The stable condition produced significantly (p ≤ 0.05) greater LyE values than the unstable condition across all muscles (Table 3). Linear descriptive statistics are presented in Table 4. In summary, mean EMG activity for the unstable condition was lower in the left and right anterior deltoids, pectorals, and triceps, and higher in the biceps. Intraclass correlation coefficients were not calculated for these data as the analyses used the entirety of the signal for the entire set performed. Furthermore, the repeatability of the data is already described nonlinearly by the LyE calculations. Because of the nature of the nonlinear calculations used (all repetitions used in 1 analysis), intraclass correlations could not be calculated between trials or repetitions. To better visualize the differences in bar movement, the marker trajectory of 1 subject was plotted and is displayed as if looking at the bencher from a sagittal view (Figure 2).

Table 1.:
Lyapunov exponent values for bar marker data, mean (SD).
Table 2.:
Sample entropy values for bar marker data, mean (SD).
Table 3.:
Lyapunov exponent values for muscle activity, mean (SD).
Table 4.:
Mean muscle activity (μV), mean (SD).
Figure 2.:
Sagittal view of bar marker during the fifth repetition of 1 subject, unstable vs. stable. *Start of the repetition.


The purpose of this study was to investigate the effects of introducing a highly unstable loading modality to a familiar, relatively constrained exercise: the bench press. To do so, we analyzed a number of nonlinear measurements to more fully describe how recreationally active individuals with little to no unstable bench press experience performed the exercise. This study was the first to investigate unstable loads resistance training using nonlinear measures. These measurements confirm some intuitive aspects of the lift and also reveal some interactions that may be useful in choosing to use this sort of training method. The main findings of this investigation were that the unstable load produced (a) more bar motion local instability in the anteroposterior and mediolateral directions than the stable condition; (b) more predictable bar motion in the vertical direction than the stable load; and (c) more locally stable muscle activation across all muscles than the stable condition.

When interpreting the data, it is useful to view the movement of the bar as the behavior of the external load and the EMG data as the control signals of the person controlling the bar. The local orbital stability calculated from the marker mounted on the end of the bar indicates how much of the available movement space that was used by the bar and how sensitive it is to small perturbations; a lower exponent indicates a lesser divergence of the reconstructed state space. In common terms, with a lower exponent the bar moved in fewer sorts of ways, whereas a higher exponent would indicate the bar moved in more ways (or with a greater number of degrees of freedom) (5). In this study, the unstable condition produced a larger exponent in the anteroposterior and mediolateral directions, but not in the primary direction of the press (Z direction) than the stable condition. This is interesting in that, as part of the person-bar system, the amount of movement space used by the bar in a vertical direction did not change despite the bar being designed deliberately to “whip” and all the resistance being added by suspending weights from latex bands.

In the anteroposterior and mediolateral directions, however, the unstable load did move in a greater variety of ways, indicating a greater sensitivity to small perturbations. This may indicate that the lifter will have to put greater effort in to stabilizing the load in the anteroposterior and mediolateral directions when lifting with this kind of load as it is more sensitive to perturbations in these directions. These findings are not surprising as the intent of this type of training is to create a scenario that is challenging to control in the directions other than that the primary motion is occurring. Our findings simply confirm that this unstable condition is providing a more challenging environment in terms of controlling the load in multiple directions.

As would be expected given the results of the LyE, measures of sample entropy determined the movement of unstable load is less predictable in the anteroposterior and mediolateral directions. This could indicate that the motion in the anteroposterior and mediolateral directions may be driven more by the intrinsic mechanical behavior of the flexible bar and the latex bands, behaving as such despite the attempts of the lifter to constrain motion as indicated through muscle activation data. In essence, although the lifter is attempting to minimize all movement of the bar except the Z direction, they are unable to fully stabilize the motion of the bar in the anteroposterior and mediolateral directions. This may be because they are inexperienced with this type of training, or it may just be an effect of this training method. Anecdotally, those with more unstable load training experience are able to better stabilize and control the motion of the bar, but it has yet to be determined whether a learning effect is possible with this type of training.

In contrast to the anteroposterior and mediolateral directions, the unstable condition is more predictable in the Z direction than the stable condition. This may be due to the unstable condition causing the muscles to be more constrained in an attempt to control the anteroposterior and mediolateral movements and “spilling over” to cause a more predictable bar path in the Z direction than the stable condition. There also exists the possibility that the vertical motion of the bar in the unstable condition caused by its inherent flexibility may be more predictable than the motion of a rigid bar during the bench press, although this seems unlikely given that the normal condition uses a rigid bar and the noise inherent in the measurement system for both conditions is similar.

When examining the EMG data, the maximal LyE may be interpreted in plain terms as the number of ways a given signal could be generated to produce the movement of the person-bar system and how sensitive that system is to perturbations. Across all muscles, the unstable lifting condition produced fewer means of organization to perform the task and was less sensitive to perturbations. In other words, lifters were more constrained in how they controlled the unstable load. This could stem from 2 aspects: first, lifting an unstable load was a novel stimulus to the lifters in this study, and it is a precept of some theories of motor learning that an amateur performing a task will restrict motion to its simplest forms when initially learning (10). However, the restriction in different motor pattern use may also reflect an effort to stabilize the unstable load bar. There may be less room for variation in the lifter's control of the bar, as slight deviances in bar motion may be amplified by the sensitivity of the unstable load condition to perturbations (as indicated by an increase in LyE for the unstable load bar motion) and perhaps lead to a failed press attempt. This also indicates that muscle activity in the unstable condition was not very sensitive to the perturbations caused by the unstable load; in other words, the lifter attempted to keep the bar moving how they intended rather than allowing the bar and weights to drive the motion.

However, during the stable condition, more degrees of freedom were used in controlling the motion of the bar. In a previous study, using the same experimental protocol, we found that integrated muscle activation was either no different between conditions or greater in the unstable condition (11). The mean EMG activity for each muscle was also calculated for this study from the same data set to eliminate the effect of the length of time required to complete each press, which was shown to be significant in the previous study. In the stable condition, the prime movers showed higher mean activity than in the unstable condition. This is to be expected given the slightly higher load and greater velocities used in the stable condition. The biceps, a biarticulate muscle group crossing both the shoulder and the elbow, showed greater mean activity in the unstable condition, however. A greater overall use of this muscle group may be the primary contributor to constricting the degrees of freedom used to lift the unstable load, acting to control motion in the elbow and shoulder and more actively transfer energy between segments.

Practical Applications

The outcomes of this study indicate that there was significant contracture of muscles not considered prime movers in the lift as a means to minimize energy loss between prime movers and bar motion. In lifting parlance, the lifters were “staying tight” in such a way to minimize the number of ways that they moved the bar during the unstable condition. This is likely an attempt to better control the bar. Training with an unstable load may be helpful for coaches trying to teach lifters and athletes to better control themselves in situations involving unstable loads (e.g., contact sports). Furthermore, the results of the EMG data suggest subjects attempted to more constrain how they activated their muscles. The practical application of these results is that the subjects attempted to “stay tighter” during the unstable loads, meaning they were less likely to activate and deactivate a muscle during the movement. Teaching athletes to “stay tight” may potentially help with preventing joint injuries in other movements and better transfer force as energy is not lost by some muscles going lax. Because the response may also be explained by the novelty of the task, and not just how lifters are attempting to control the bar, it would be useful in future studies to have people experienced with an unstable benching exercise perform the task and compare whether similar control strategies are used.


This project was funded by Westbrook College of Health Professions Summer Undergraduate Research Fellowship Grant. There are no conflicts of interested associated with this study. The results of this study do not constitute endorsement of the product by the authors or the National Strength and Conditioning Association.


1. Baechle TR, Earle RW; National Strength and Conditioning Association (U.S.). Essentials of Strength Training and Conditioning. Champaign, IL: Human Kinetics, 2008.
2. Dingwell JB, Smallwood RF, Cusumano JP. Trial-to-trial dynamics and learning in a generalized, redundant reaching task. J Neurophysiol 109: 225–237, 2013.
3. Dunnick DD, Brown LE, Coburn JW, Lynn SK, Barillas SR. Bench press upper-body muscle activation between stable and unstable loads. J Strength Cond Res 29: 3279–3283, 2015.
4. Gates DH, Dingwell JB, Scott SJ, Sinitski EH, Wilken JM. Gait characteristics of individuals with transtibial amputations walking on a destabilizing rock surface. Gait Posture 36: 33–39, 2012.
5. Kang HG, Dingwell JB. Dynamics and stability of muscle activations during walking in healthy young and older adults. J Biomech 42: 2231–2237, 2009.
6. Kennel MB, Brown R, Abarbanel HD. Determining embedding dimension for phase-space reconstruction using a geometrical construction. Phys Rev A 45: 3403–3411, 1992.
7. Kohler JM, Flanagan SP, Whiting WC. Muscle activation patterns while lifting stable and unstable loads on stable and unstable surfaces. J Strength Cond Res 24: 313–321, 2010.
8. Lawrence M, Carlson L. Effects of an unstable load on force and muscle activation during a parallel back squat. J Strength Cond Res 29: 2949–2953, 2015.
9. McBride JM, Larkin TR, Dayne AM, Haines TL, Kirby TJ. Effect of absolute and relative loading on muscle activity during stable and unstable squatting. Int J Sports Physiol Perform 5: 177–183, 2010.
10. Newell K. Change in motor learning: A coordination and control perspective. Mortiz 9: 1–6, 2003.
11. Ostrowski S, Carlson LA, Lawrence MA. Effect of an unstable load on primary and stabilizing muscles during the bench press. J Strength Cond Res 31: 430–434, 2017.
12. Roos PE, Dingwell JB. Influence of simulated neuromuscular noise on the dynamic stability and fall risk of a 3D dynamic walking model. J Biomech 44: 1514–1520, 2011.
13. Wolf A, Swift J, Swinney H, Vastano J. Determining Lyapunov exponents from a time series. Physica 16D: 285–317, 1985.

Lyapunov exponent; sample entropy; bandbell bar

© 2016 National Strength and Conditioning Association