Cable and pulley systems have been incorporated into selectorized and plate-loaded resistance training machines since their inception, especially in exercises where a change in the direction of an applied force is required to move the resistance, as in the classic lat pull-down or triceps extension push-down exercises. With the increased interest in functional training, there has been a resurgent interest in cable training systems and a concomitant increase in the production and evolution of these machines. Although cables are part of the linkages within many resistance training machines, cable training machines, where the cables and handles constitute the human with machine interface, are now gaining popularity (2,11,15). The clinical relevance of these machines to functional training is supported by the concept of movement specificity. Unlike plate-loaded and selectorized machines, cable machines allow a considerably greater number of degrees of freedom and in most cases require considerably greater activation of the core musculature as force is sequentially transferred in a kinetic chain from the lower to upper body.
Many activities of daily living (ADL), such as transfer of objects from one surface to another, sweeping, raking, or retrieving objects from the floor, require the transfer of force from the lower to upper body musculature through the core muscles. In contrast, many other activities may be upper or lower body dominant, such as washing dishes or climbing stairs, respectively. Although the former may be addressed by multi-joint exercises using specific kinetic chains, the latter may be better addressed by more isolated movement patterns. This same training dichotomy may also be applied to sports activities. For example, activities such as hitting a tennis groundstroke, driving a golf ball, or throwing a punch demand substantial force and velocity transfers from the lower to upper body, whereas other sports such as distance running or the bench press during powerlifting competition are predominantly dependent on the lower and upper body, respectively.
In a study that compared cable training with more traditional resistance training modalities, Santana et al. (15) noted that the activation of the chest and shoulder muscles during cable training was dependent on the levels of stabilization afforded by the lower body and core muscles, limiting the levels of muscle activation during the cable chest press compared with the traditional bench press.
In addition to the muscle utilization patterns noted above, a second biomechanical consideration associated with targeted training prescription is joint angle specificity. A number of studies support the importance of angle specificity during both isometric and dynamic training. The majority of the isometric training studies have reported that the greatest training responses occurred at the training angle; however, increases were seen across a rather broad range of motion (ROM) (10,16). The studies that have examined angle and ROM specificity during isoinertial training are more limited. An early study by Graves et al. (5) showed angle-specific training responses during limited ROM training using variable resistance training equipment. In a later study comparing variable resistance with cable training (12), variable resistance training allowed a greater ROM and muscle activation for the rotator cuff as resistance increased; however, these data were collected using restricted elbow and shoulder angles during cable training and the percentage loads were dictated by the maximums (1 repetition maximum, 1RM) produced during the cable exercise alone rather than relative to each exercise.
Angle specificity has also been evidenced in sport. Herzog et al. (7) reported significant differences in isometric knee extension length-moment relationships between cyclist, runners, and speed skaters. Additionally, a study by Rousanoglou et al. (14) reported angle-specific differences in knee extension torque between younger (≤16.49 years) and older (≥16.50 years) girls dictated by the sports in which they participated. Finally, when examining specificity as it relates to ADL performance, Kerr et al. (8) noted that recumbent cycling matched both muscle utilization patterns and ROM of the step-up and sit-to-stand tests.
Given the variations in muscle utilization patterns and kinematics seen because of specific training modalities, and the implications of using cable vs. plate-loaded or selectorized machines when designing effective interventions for improving ADL or sport performances, this study analyzed muscle utilization patterns and kinematics during the performance of similar exercises using cable and selectorized machine training. We hypothesized greater core and accessory muscle activity for each exercise tested during cable training because of the need to stabilize specific body segments during each exercise. In contrast, we reasoned that greater activity levels would be seen in the specific muscle targeted by each exercise during selectorized training because of greater stabilization and isolation of those muscle groups.
Experimental Approach to the Problem
Because specific sports activities and ADL may involve isolated movements or movements requiring sequential muscle firing of selected kinetic chains, this study examined muscle firing patterns and kinematics during the performance of comparative exercises during cable-based strength training vs. strength training using selectorized weight machines commonly used during training interventions. A series of repeated-measures analysis of variance (ANOVA) was used to examine the differences among muscle activation patterns and kinematics between these 2 conditions. Participants completed 5 repetitions of overhead press, biceps curl, and chest press exercises with 8 repetition maximum (8RM) loads on a cable-based strength training machine and selectorized plate machines in a randomized order. Muscular activities of the pectoralis major (PM), anterior deltoid (AD), biceps brachii (BB), rectus abdominis (RA), external oblique, and triceps brachii (TB) were continuously assessed using a wireless surface electromyography (EMG) unit, whereas shoulder, elbow, hip, and knee joint angles were captured using a high-speed camera and quantified using 2-dimensional kinematic analyses. The results of this study can serve as an initial evaluation of how these machines can vary muscle utilization levels and patterns of movement allowing them to be used selectively in targeted exercise interventions.
A total of 15 participants (9 men, 6 women; mean age ± SD, 24.33 ± 4.88 years; age range 18–38) were recruited on a voluntary basis through personal contacts from an opportunity sample of recreationally active students in a university research program. The initial criterion for inclusion into the study was that the individual had no history of neuromuscular disease or injury that might affect his or her performance during the testing session. All participants were informed of experimental procedures and signed a written consent approved by the University's Subcommittee for the Use and Protection of Human Subjects. Table 1 contains the characteristics for the subjects who participated in this study.
Testing was performed across 2 sessions separated by at least 24 hours but no longer than 48 hours. Subjects were instructed to refrain from exercise for 24 hours before testing; however, no control was exerted on the time of day at which testing sessions were performed. During the first testing session, 8RM loads were established for the bicep curl, chest press, and overhead press on both a cable-based strength machine (Cybex Bravo Pro; Cybex International, Inc., Medway, MA, USA) and selectorized arm curl, overhead press, and chest press (VR2; Cybex International, Inc.). These exercises were chosen because of their regular use during recreational and fitness-based programs. Participants used a neutral grip on the selectorized plate machines and replicated this grip on the cable machine. To determine the 8RM load, participants began with a 10-repetition light weight warm-up set. After a 1-minute rest, the weight was increased to an estimated 8RM and 8 repetitions were attempted. If participants successfully completed the 8 repetitions, loads were increased until the true 8RM in each of the 6 resistance training exercises was achieved. A 2-minute recovery was provided between sets of the same exercise, and a 5-minute recovery was given between exercises. The true 8RM for each exercise was determined in 3–5 sets.
On the second testing day, after a short warm-up, participants performed 5 repetitions using the 8RM of each exercise: cable machine biceps curl, cable machine overhead press, cable machine chest press, selectorized plate machine biceps curl, selectorized plate machine overhead press, and selectorized plate machine chest press. The order of exercises was randomized among subjects. Subjects performed the repetitions to a metronome set at 1 second per contractile stage (concentric, eccentric). Five-minute recovery periods were provided between trials.
Electromyography was used to quantify neuromuscular activity in the 6 muscles of interest: PM, AD, BB long head, RA, external oblique, and TB lateral head. A bipolar surface configuration was used to maximize the reception area while controlling the potential for cross-talk among the muscles examined. The skin overlying each muscle was shaved, abraded, and cleansed with rubbing alcohol to remove dead surface tissues and oils and reduce skin impedance. Disposable Ag/AgCl dual electrodes (Noraxon USA, Inc., Scottsdale, AZ, USA) were then positioned parallel with the underlying muscle fibers according to Cram's Introduction to Surface Electromyography recommendations (3). To reduce movement artifacts, the electrodes were held against the skin using a self-adherent wrap. Raw EMG signals were recorded using a wireless EMG telemetry system (Noraxon TeleMyo Direct Transmission System; Noraxon USA, Inc.) with an input impedance of 100 MΩ and a common mode rejection ratio of 100 db. The gain was set at 2,000 Hz with band pass filtering set between 3 and 500 Hz. Signals were sampled at a frequency of 1,500 Hz, digitized using a 16-bit A/D converter (Noraxon USA, Inc.) and stored on a personal computer. Recorded EMG signals from each muscle were analyzed using a custom-built software program (LabView; National Instruments Corporation, Austin, TX, USA) written to quantify the rmsEMG. The root mean square of the EMG (rmsEMG) signal was used as a measure of average muscle activation for each muscle during each resistance exercise.
A JVC GC-PX100 High-Speed Camera (JVC Corporation, Long Beach, CA, USA) was used to measure the starting angles, ending angles, and ROM for the elbow and shoulder joints during the biceps curl; hip and knee during the chest press; and hip, knee, and shoulder during the overhead press. Because all measurements were made solely in the sagittal plane, evaluations were limited to these specific joints for each exercise. Video footage was downloaded to our laboratory computers and Kinovea software (Version 0.8.15) was used to quantify each variable.
Separate 6 (muscle) × 2 (machine) repeated-measures mixed ANOVA were used to evaluate comparative rmsEMG levels among muscles for the 3 exercises. For kinematic variables, separate 2 (joint) × 2 (machine) mixed ANOVA were used to evaluate biceps curl and chest press and a 3 (joint) × 2 (machine) mixed ANOVA for the overhead press. Statistical significance was set a priori at p ≤ 0.05. All statistical procedures were performed using the SPSS statistical package (version 22.0; SPSS, Inc., Chicago, IL, USA). When a significant main effect or interaction was detected, the least significant difference (LSD) post hoc test was used to determine the sources of the differences.
The mean values and standard errors for the rmsEMG values of individual muscles during the 3 exercises are provided in Tables 2–4. As shown in Table 2 and Figure 1, the biceps curl exercise produced a significantly higher mean rmsEMG for the PM (Mdiff ± SE = 68.99 ± 25.55 μV; 95% confidence interval [CI] = 14.18–123.79 μV; p = 0.017) and the AD (Mdiff ± SE = 77.31 ± 24.10 μV; 95% CI = 25.62–129.01 μV; p = 0.006) on the cable vs. the plate-loaded machine.
Furthermore, as shown in Table 3 and Figure 2, the chest press exercise produced significantly higher mean rmsEMG on the cable than the plate-loaded machine for the BB (Mdiff ± SE = 120.24 ± 26.59 μV; 95% CI = 63.21–177.27 μV; p < 0.00011), RA (Mdiff ± SE = 39.78 ± 17.47 μV; 95% CI = 2.30–77.27 μV; p = 0.039), and external oblique (Mdiff ± SE = 16.05 ± 5.93 μV; 95% CI = 3.33–28.17 μV; p = 0.017). In contrast, the TB showed greater activity levels for the plate machine vs. cable (Mdiff ± SE = 157.35 ± 31.33 μV; 95% CI = 224.55–90.15 μV; p < 0.0001).
Finally, for the overhead press exercise, significant differences were detected between conditions for the BB (Mdiff ± SE = 117.32 ± 25.82 μV; 95% CI = 61.94–172.70 μV; p < 0.0001) and external obliques (EO; Mdiff ± SE = 11.60 ± 3.82 μV; 95% CI = 3.41–19.79 μV; p = 0.009) favoring the cable over the plate-loaded machine. Comparisons can be seen in Table 4 and Figure 3.
The mean values and standard errors for the starting angle, ending angle, and ROM of the elbow and shoulder joints recorded during the biceps curl are provided in Table 5. As shown in the table, there were significant differences between the starting angles of the elbow joint for the cable biceps curl exercise vs. the plate-loaded biceps curl machine (Mdiff ± SE = 7.45 ± 3.44°; 95% CI = 0.57–14.32°; p = 0.036). There were also significant differences for the ending angles of the elbow joint between the 2 machines (Mdiff ± SE = 6.88 ± 1.73°; 95% CI = 3.44–10.89°; p = 0.001). However, when comparing the ranges of motion for the elbow joint on the 2 machines, there were no significant differences (Mdiff ± SE = 2.86 ± 5.16°; 95% CI = −7.45 to 13.18°; p = 0.585). There were also significant differences between the starting angles of the shoulder joint on the cable vs. the plate-loaded machine (Mdiff ± SE = 54.93 ± 17.19°; 95% CI = 51.00–58.44°; p < 0.0001). Additionally, there were significant differences between the machines when comparing the ending angles of the shoulder joint (Mdiff ± SE = 28.65 ± 3.04°; 95% CI = 22.92–34.95°; p = 001). Finally, when comparing the ranges of motion of the shoulder joint between the machines, a significant difference was detected (Mdiff ± SE = 39.63 ± 2.29°; 95% CI = 34.95–44.12°; p < 0.0001).
Results for the starting angles, ending angles, and ranges of motion of the hip and knee joints associated with the chest press are shown in Table 6. The table shows that the starting angles of the hip joint on the cable machine were significantly higher than the starting angles on the plate-loaded machine (Mdiff ± SE = 37.24 ± 2.69°; 95% CI = 31.74–42.74°; p < 0.0001). Similarly, there were significant differences between the ending angles of the hip joint on the 2 machines (Mdiff ± SE = 25.15 ± 2.64°; 95% CI = 19.82–30.54°; p < 0.0001). The ranges of motion for the hip joint were also significantly larger on the cable machine compared with the plate machine (Mdiff ± SE = 8.88 ± 1.03°; 95% CI = 6.82–10.94°; p < 0.0001). During the chest press, the angles measured at the start of the repetition were significantly lower on the plate-loaded vs. the cable machine (Mdiff ± SE = 71.62 ± 3.04°; 95% CI = 65.43–77.75°; p < 0.0001). There were also significant differences between the cable and plate machine for the ending angles of the knee joint (Mdiff ± SE = 75.00 ± 2.75°; 95% CI = 69.39–80.62°; p < 0.0001). Finally, the ranges of motion of the knee joint on the cable machine were significantly higher than those of the plate-loaded machine (Mdiff ± SE = 4.47 ± 0.92°; 95% CI = 2.64–6.30°; p < 0.0001).
Table 7 provides the mean values and standard errors for the starting angles, ending angles, and ranges of motion of the hip, knee, and shoulder joints for the overhead press exercise. The table shows significant differences between the starting angles of the hip joint on the cable machine vs. the plate-loaded machine (Mdiff ± SE = 48.32 ± 2.46°; 95% CI = 44.69–54.77°; p < 0.0001). The analyses of the ending angles for the hip joint also showed significant differences between the 2 machines (Mdiff ± SE = 52.71 ± 2.24°; 95% CI = 49.85–57.41°; p < 0.0001). A comparison of the ranges of motion of the hip joints on the 2 machines showed significant differences as well (Mdiff ± SE = 2.12 ± 0.63°; 95% CI = 0.86–3.38°; p = 0.002). For this exercise, there were also significant differences between the starting angles of the knee joint on the 2 machines (Mdiff ± SE = 7.28 ± 3.61°; 95% CI = 65.15–79.93°; p < 0.0001). When comparing the ending angles of the knee joint, there were also significant differences between the 2 machines (Mdiff ± SE = 80.67 ± 3.00°; 95% CI = 74.66–86.63°; p < 0.0001). The ranges of motion of the knee joint were significantly different between the 2 machines (Mdiff ± SE = 6.13 ± 1.26°; 95% CI = 3.55–8.65°; p < 0.0001). When comparing the starting angles of the shoulder joint on the overhead press, no significant differences were evident between the 2 machines, although a trend was evident (Mdiff ± SE = 16.90 ± 8.42°; 95% CI = −0.69 to 34.44°; p = 0.059). There were significant differences between the ending angles of the shoulder joint between the 2 machines (Mdiff ± SE = 16.85 ± 2.92°; 95% CI = 10.77–22.86°; p < 0.0001). Finally significant differences were found between the ranges of motion of the shoulder joint on the 2 machines (Mdiff ± SE = 73.17 ± 4.58°; 95% CI = 63.71–82.62°; p < 0.0001) (Figures 4A, B, 5A, B, 6A, B).
The major finding of this study was that the activities of selected muscles during comparative exercises varied by machine use as did beginning and ending angles and ROM for specific joints. In examining muscle activity levels, it should be noted that the differences recorded between machines were seen primarily in accessory rather than the muscles commonly targeted during each exercise.
For the arm curl exercises, similar firing patterns were seen for the biceps and TB, RA, and EO. In contrast, the activity levels of the PM and AD were significantly higher for cable vs. plate machine training. These results are somewhat surprising given the greater loads used during the plate-loaded exercise and the expectation of increased EMG activity with increasing load for any selected muscle groups during the biceps curl (6,13). However, there is a precedent set for varying EMG responses in muscle activities of the BB with different arm curl exercises that supports our findings. Oliveira and Gonçalves (11) reported that the electrical activity of the BB was greater during a standing biceps curl compared with a seated biceps curl. These researchers argued that the seated position affords a mechanical position associated with lower levels of fatigue and therefore reduced EMG activity that may have decreased the differences in rmsEMG between the 2 exercises. The lack of difference in biceps activity between machines is further explained by the small, though significant, variabilities in starting and ending angles and, moreover, the similarity in ROM for both machines. In contrast, the significant differences seen between machines for the starting angle and ending angle of the shoulder and the significantly greater ROM for the cable machine may explain the greater activity levels of the AD and PM. When examining the illustrations for each exercise (Figures 4A, B), it is quite obvious that the support pad of the arm curl machine stabilized the upper arm while subjects attempted to maintain upper arm positions using the PM and AD as shoulder stabilizers during the cable lift.
For the chest press, the cable and plate-loaded machines again produced no significant differences in muscle activities for 2 of the major muscles targeted by the exercise, the PM and AD. Although analysis of the shoulder and elbow joint ankles was not possible because of the camera view being limited to the sagittal plane, examination of the photographs and illustrations in Figures 5A, B shows the similar patterns at the shoulder joint for both exercises. In contrast, the TB, a muscle group targeted by this exercise, did show greater activity during plate-loaded training compared with cable training. One explanation for this pattern is the variation in resistance between the 2 exercises (cable, 13.1 ± 4.1 kg; plate, 63.2 ± 31.5 kg). Because the greater level of muscle activity for the PM did not reach statistical significance (cable, 213.96 ± 39.54 μV; plate, 264.49 ± 63.82 μV), the need to move over a fivefold greater load may have been achieved through a compensatory level of firing of the TB group (cable, 244.89 ± 34.05 μV; plate, 402.24 ± 48.50 μV) along with the high level of stabilization provided by the plate-loaded machine's seatback. The differences in training loads and muscle activation patterns seen in the current study are reflective of those seen in a similar study performed by Santana et al. (15); however, unequivocal comparison between the 2 studies is limited by their study design, which did not offer a direct comparison of muscle activation patterns between machines and used a single arm rather than bilateral cable press. The higher activity levels for the RA and EO during cable vs. plate machine training is expected because of the need to stabilize the trunk during the standing cable exercise vs. the external stability offered by the cable machine seatback. Although, as noted above, Santana et al. (15) did not compare cable chest press and bench press, the patterns of relative activity among muscles within each exercise confirm the greater activity levels for the RA and EO seen in the current study. Our results are also supported by the results reported by Calatayud et al. (2), who examined the muscle activities of the RA, EO, PM, AD, TB, upper trapezius, serratus anterior, and posterior deltoid during push-ups on different suspension systems, floor push-ups with and without elastic resistance, and bench press or standing cable press exercises performed at 50, 70, and 85% 1RM in college-aged men. They reported that elastic-resisted push-ups induced similar EMG patterns in the prime movers as the bench press at high loads while also providing a greater core challenge. Additionally, they noted that suspended push-ups are a highly effective way to stimulate the abdominal muscles. They concluded that the PM, AD, and serratus anterior showed their highest levels of activity in the bench press, which afforded a more stable pushing condition, whereas the abdominal muscles, TB, posterior deltoid, and upper trapezius showed greater activity under the less stable environments afforded by the suspended and cable press conditions.
Finally, for our last exemplar, the overhead press showed no differences between machines in the prime muscles, including the AD, TB, and accessory muscles, including the PM and RA. There were significantly higher activity levels, however, during cable training for the BB and EO. The greater activation of the BB group may be attributable to its function in assisting the movement of the arms upward and forward at the shoulder and in sideward stabilization, and the fact that these functions were performed through a greater ROM during the cable overhead press (172.4 ± 3.3°) than the cable overhead press (99.3 ± 3.5°). The angles at the knee (starting, 168.6 ± 2.1°; ending, 170.1 ± 1.5°) and hip (starting, 172.6 ± 1.2°; ending, 172.4 ± 1.0°), each approaching 180° are associated with the standing position during the cable overhead press, whereas the greater ROMs at each joint (knee, 3.9 ± 0.7°; hip, 7.9 ± 1.3°) are indicative of force transfer from the lower to the upper limbs during the exercise (Figures 6A, B). It can be logically maintained that, given the function of the EO in compression of the abdominal wall and increasing intra-abdominal pressure, their activity levels should be greater during the cable vs. plate-loaded machine (1,4,9).
Our results demonstrate that the use of cable-based vs. selectorized machines can produce very different muscle utilization patterns even when the same basic exercises are performed. This result is not unexpected because both joint angles and ROMs varied considerably during the same exercises performed on each type of machine. Because the levels of activity can be varied by the machine used during the same exercise, machine choice becomes a viable tool that can be added to the coach, therapist, or personal trainer's repertoire when providing targeted exercise. For example, the higher activation levels of the core muscles during the chest press and overhead press exercises during cable vs. selectorized machine use indicate that cable machines may be more effective when targeting sport and ADL activities that depend heavily on serape-dominated movements (transitions using rotational movements that transfer force from the lower to upper body through the core). Common ADL relying heavily upon such patterns include sweeping or object transfer from one counter to another, while groundstrokes in tennis or driving a golf ball are typical examples in sport. In contrast, the higher activity levels seen in the biceps and AD during the arm curl when using plate-loaded machines indicate that this type of training may prove a more effective choice when targeting activities where more linear movement patterns using limited body segments dominate. Examples include placing an object on an overhead shelf or scrubbing a countertop during ADL; or in sports, targeting specific muscles during bodybuilding or paddling during kayaking. Future studies should use other exercises, examine patterns of activity across a greater number of muscles, and use more sophisticated movement analysis techniques to increase the scope of this research and expand our understanding of machine choices as they relate to specific prescriptive strategies.
1. Arjmand N, Shirazi-Adl A. Role of intra-abdominal pressure in the unloading and stabilization of the human spine during static lifting tasks. Eur Spine J 15: 1265–1275, 2006.
2. Calatayud J, Borreani S, Colado JC, Martin F, Rogers ME. Muscle activity levels in upper-body push exercises with different loads and stability conditions. Phys Sportsmed 42: 106–119, 2014.
3. Criswell E. Cram's Introduction to Surface Electromyography. Boston, MA: Jones and Bartlett Publishers, 2011.
4. Freeman S, Karpowicz A, Gray J, McGill S. Quantifying muscle patterns and spine load during various forms of the push-up. Med Sci Sports Exerc 38: 570–577, 2006.
5. Graves JE, Pollock ML, Jones AE, Colvin AB, Leggett SH. Specificity of limited range of motion variable resistance training. Med Sci Sports Exerc 21: 84–89, 1989.
6. Gupta V, Suryanarayanan S, Reddy NP. Fractal analysis of surface EMG signals from the biceps. Int J Med Inf 45: 185–192, 1997.
7. Herzog W, Guimaraes AC, Anton MG, Carter-Erdman KA. Moment-length relations of rectus femoris muscles of speed skaters/cyclists and runners. Med Sci Sports Exerc 23: 1289–1296, 1991.
8. Kerr A, Rafferty D, Moffat F, Morlan G. Specificity of recumbent cycling as a training modality for the functional movements; sit-to-stand and step-up. Clin Biomech (Bristol, Avon) 22: 1104–1111, 2007.
9. McGill S, Andersen J, Cannon J. Muscle activity and spine load during anterior chain whole body linkage exercises: The body saw, hanging leg raise and walkout from a push-up. J Sports Sci 33: 419–426, 2015.
10. Morrissey MC, Harman EA, Johnson MJ. Resistance training modes: Specificity and effectiveness. Med Sci Sport Exerc 27: 648–660, 1995.
11. Oliveira AS, Goncalves M. Positioning during resistance elbow flexor exercise affects electromyographic activity, heart rate, and perceived exertion. J Strength Cond Res 23: 854–862, 2009.
12. Peltonen H, Arokoski J, Kallinen M, Pullinen T. Muscle loading and activation of the shoulder joint during humeral external rotation by pulley and variable resistance. J Electromyogr Kinesiol 22: 424–430, 2012.
13. Roman-Liu D, Tokarski T. Upper limb load as a function of repetitive task parameters: Part 2—An experimental study. Int J Occup Saf Ergon 11: 103–112, 2005.
14. Rousanoglou E, Nikolaidou M, Boudolos K. Discrimination of young women athletes and nonathletes based on anthropometric, jumping, and muscular strength measures 1, 2. Percept Mot Skills 102: 881–895, 2006.
15. Santana JC, Vera-Garcia FJ, McGill SM. A kinetic and electromyographic comparison of the standing cable press and bench press. J Strength Cond Res 21: 1271–1277, 2007.
16. Weir JP, Housh DJ, Housh TJ, Weir LL. The effect of unilateral eccentric weight training and detraining on joint angle specificty, cross-training, and the bilateral deficit. J Orthop Sports Phys Ther 22: 207–213, 1995.