Primary and submovement control of aiming in C6 tetraplegics following posterior deltoid transfer
© Robinson et al.; licensee BioMed Central Ltd. 2014
Received: 20 January 2014
Accepted: 21 July 2014
Published: 23 July 2014
Upper limb motor control in fast, goal-directed aiming is altered in tetraplegics following posterior-deltoid musculotendinous transfer. Specifically, movements have similar end-point accuracy but longer duration and lower peak velocity than those of age-matched, neurotypical controls. Here, we examine in detail the interplay between primary movement and submovement phases in five C6 tetraplegic and five control participants.
Aiming movements were performed in two directions (20 cm away or toward), with or without vision. Trials that contained a submovement phase (i.e., discontinuity in velocity, acceleration or jerk) were identified. Discrete kinematic variables were then extracted on the primary and submovements phases.
The presence of submovements did not differ between the tetraplegic (68%) and control (57%) groups, and almost all submovements resulted from acceleration and jerk discontinuities. Tetraplegics tended to make a smaller amplitude primary movement, which had lower peak velocity and greater spatial variability at peak velocity. This was followed by a larger amplitude and longer duration secondary submovement. Peak velocity of primary movement was not related to submovement incidence. Together, the primary and submovement phases of both groups were equally effective in reducing end-point error.
C6 tetraplegic participants exhibit some subtle differences in measures of motor behaviour compared to control participants, but importantly feedforward and feedback processes work effectively in combination to achieve accurate goal-directed aiming.
Tetraplegics with a spinal cord lesion at C6 typically have normal use of the biceps brachii and wrist extensor muscles but paralysis of the triceps brachii, wrist flexors and all finger and thumb muscles. The functional imbalance at the elbow between the biceps brachii and triceps brachii means that there is no antagonist to elbow flexion or agonist for elbow extension. This has consequences for elbow flexion movements, with tetraplegics exhibiting greater movement amplitudes, as well as longer acceleration and deceleration times than unimpaired controls. Other studies, however, have shown that the aiming movements of tetraplegics retain many kinematic similarities to the movement of neurotypical control participants[2, 3].
A solution to permanently replace the function of the paralysed triceps brachii is posterior-deltoid (PD) musculotendinous transfer. Recent evaluation of the effects of PD transfer on aiming movements with the upper limb showed that typical spatial and temporal characteristics are not completely restored compared to age-matched, neurotypical control participants. Specifically, although tetraplegics who had undergone PD transfer took no more time than controls to plan and initiate aiming movements, and were no more reliant on vision for online regulation, they maintained overall accuracy by reducing movement velocity and lengthening movement time. It was suggested that tetraplegic participants had learned to adopt a feedforward and feedback control that minimizes the spatial variability associated with a noisier neuromuscular system; for a discussion of strategic influences on motor control see[6, 7]. In other words, tetraplegics exhibited longer movement time because they recognized moving at a faster speed similar to neurotypical control participants would produce more variable trajectories[8, 9], which would then require online correction. Additional support for this interpretation was evident in the relatively coarse classification (i.e., zero crossing in velocity or acceleration profile) and analysis of submovements (i.e., total number), which revealed no group differences.
The aim of the current study was to describe in more detail the primary movement and submovement phases of discrete adapted aiming movements performed by tetraplegics with PD transfer. Extending upon previous work where all trials were considered, here we reanalysed only those trials containing a primary and submovement phase (>50%). In addition to measures of movement kinematics, we quantified the frequency and distribution (velocity, acceleration or jerk discontinuity) of submovements, as well as their amplitude and duration. We also sought to determine if peak velocity is related to the incidence of submovements on a trial-by-trial basis, and finally if submovements reduced end-point error. Aiming movements were performed with a ball transfer unit that enabled participants to perform a ‘frictionless’ multi-directional aiming movement away from or toward the body. Accordingly, we were able to examine if performing aiming movements that required a different coordinated recruitment of the agonist and antagonist muscles impacted the control of primary movement and submovement phases. In addition, the discrete aiming task was performed while wearing liquid crystal goggles such that on half the trials participants were prevented from seeing either their hand or the target during the execution of the aiming movement. In this way, we were able to investigate the influence of having access to visual and/or proprioceptive feedback. Finally, by comparing these data to those exhibited by age-matched, neurotypical controls, it was our intention to provide further insight into the motor control strategies of tetraplegics with PD transfer.
Materials and methods
Details of injury classification and surgical history for the tetraplegic participants
Level of injury
Time elapsed since
All participants were seated and the tetraplegic participants remained in their own wheelchair. All participants had trunk movement restricted with a chest strap to avoid trunk motion and thereby facilitate the contribution of elbow motion to task completion. All participants completed a total of 64 horizontal aiming movements, which were separated into four blocks (2 vision, 2 no vision) of 16 trials. Within each block, target location was randomized so that each participant aimed equally away from or towards the body. Blocks were counterbalanced across participants, and advance information was provided regarding availability of vision. Each aiming movement began with the participant pressing down on the central “home” switch with the ball transfer unit, after which there was a random foreperiod between 500–750 ms, followed by an audio cue indicating aiming direction and that the trial should commence. A 400 Hz tone indicated that the participant should aim in the away direction whereas a 200 Hz tone indicated the toward direction. For the trials without vision, the glasses instantaneously became opaque for 5 s once the central “home” switch was released. Participants understood that they were required to perform the aiming task without vision and keep their arm in their final position until the glasses became transparent again. Thus they received terminal visual feedback but not concurrent visual feedback about their aiming. For each combination of condition and aiming direction, participants were required to move the arm as fast and as accurately as possible in order to locate the ball transfer unit on the target. No external feedback was given regarding movement time or final accuracy of the ball-transfer unit relative to the target. Importantly, while the ball transfer unit obscured vision of the target when it was in close proximity, this situation is no different from aiming with the fingertip or a tool to small targets. Also, participants were familiarised with the ball transfer unit and thus the location of the single large ball relative to the wrist-joint centre.
All data were extracted using Visual 3D (v4.00.20, C-Motion, Germantown, MD, USA) and custom Matlab scripts. The position data from the wrist marker were passed through a Butterworth fourth-order low pass (10 Hz) filter, after which velocity, acceleration and jerk of the wrist marker was calculated using a 2-point finite difference algorithm. A custom-written routine implemented in Matlab then identified movement start and end, which were defined as the moment wrist velocity exceeded, and then fell below, 20 mm/s for 150 ms. Movement time was defined as the interval between movement start and end. Having next identified initial peak in the velocity profile, the routine identified if one of the following events occurred prior to actual end of movement: 1) zero crossing in velocity (i.e., movement reversal); or 2) zero crossing in acceleration; or 3) zero crossing in jerk. The routine searched for these events in the prescribed order, and thus a zero crossing in acceleration or jerk would not be identified if it followed a zero crossing in velocity. If none of the above criteria was met, the aiming attempt was deemed to be completed with a primary movement alone and was not subjected to further analysis. Conversely, if a criterion was met, the trial was deemed to contain a submovement (i.e., movement occurring from the end of the primary movement until movement end), which was classified according to the zero crossing identified; velocity = type 1, acceleration = type 2, jerk = type 3. No search was made for consecutive discontinuities and thus the number of submovements within a trial.
Trials containing a primary movement and submovement were used to calculate a number of dependent variables: percentage of trials with a submovement, percentage distribution of types of submovement (i.e., Type 1, 2 or 3), amplitude of the primary movement (mm) and secondary submovement in the main axis (mm), submovement duration (ms), peak velocity (mm/s), peak acceleration (mm/s2), and spatial variability at peak velocity (i.e., standard deviation of the wrist displacement). To test the hypothesis that speed of aiming movement on a trial-by-trial basis may account for the presence of submovements, the relationship between peak velocity of the primary movement and the presence of secondary submovements (yes = 1, no = 0) was examined using a bi-serial correlation for each participant. Finally, to determine if the secondary submovement was functional, we also calculated the number of trials in which: 1) absolute error at the end of the primary movement was greater than absolute error at movement completion (functional); 2) absolute error at the end of the primary movement was less than absolute error at movement completion (non-functional).
Where appropriate, these performance and kinematic variables were analysed in SPSS (v.18, IBM, Armonk, NY, USA) using separate 2 group (tetraplegic, control) x 2 condition (vision, no vision) x 2 direction (away from the body, towards the body) mixed analyses of variance with repeated measures on the last two factors. Post hoc tests were conducted using the Tukey HSD procedure. To ensure normality of distribution, correlation coefficients were converted to Fisher Z scores and percentage data were subjected to arcsine transformation. Alpha was set at P < 0.05.
Mean (±standard deviation) kinematic variables for control and tetraplegic participants as a function of condition and direction (VIS A - Vision Away; NOV A - No Vision Away; VIS T - Vision Towards; NOV T - No Vision Towards)
Movement time (ms)
Peak Velocity (mm/s)
Peak Acceleration (mm/s2)
Primary Movement Amplitude (mm)
Submovement Amplitude (mm)
Submovement Duration (ms)
Spatial Variability at Peak Velocity (mm)
For primary movement amplitude, there was a significant main effect of direction [F(1,8) = 6.07, P < 0.05], as well as a condition × direction interaction [F(1,8) = 5.50, P < 0.05]. Post-hoc testing showed that primary movement amplitude was greater when aiming away from (188 mm) than toward (169 mm) the body, and that this difference was somewhat larger in the vision than no vision condition (see Table 2). The main effect of group for primary movement amplitude approached conventional significance, with tetraplegics tending to make smaller amplitude primary movements than controls (172 mm vs. 185 mm, F(1,8) = 4.30, P = 0.07). Analysis of submovement amplitude revealed a significant main effect of group, with tetraplegics making larger corrections than controls (25 vs. 14 mm; F(1,8) = 6.01, P < 0.05). For duration of corrective submovements, the main effect of group also approached significance, with tetraplegics tending to exhibit a longer correction time than controls (185 ms vs. 123 ms; F(1,8) = 3.98, P = 0.08). These effects indicate that the tetraplegic group tended to make a smaller amplitude primary movement, which was followed by a larger amplitude and longer duration secondary submovement.
For magnitude of spatial variability at peak velocity in submovement trials there was a significant main effect of group [F(1,8) = 6.232, P < 0.05], and significant condition × direction × group interaction [F(1,8) = 6.781, P < 0.05]. Overall, the tetraplegic group (17 mm) had greater spatial variability at peak velocity than the control group (10 mm). Also, while spatial variability at peak velocity in the control group was consistent across vision conditions and aiming direction, the tetraplegic group exhibited greatest variability when aiming away from the body with no vision and towards the body with vision (Table 2).
The results of the bi-serial correlation indicated that across each combination of vision and direction, there were no significant within-participants relationships for the tetraplegic group (grand mean Fisher Z score = 0.01) or control group (grand mean Fisher Z score = −0.33). The same pattern of results was found for bi-serial correlations between peak acceleration of the primary movement and the presence of submovements (tetraplegic group: grand mean Fisher Z score = 0.03; control group: grand mean Fisher Z score = −0.26). Therefore, having shown that peak velocity of the primary aiming movement within a trial was not related to the incidence of a submovement, we sought to determine if the submovements were functional in reducing end-point error. To this end, we calculated the percentage of trials in which: 1) absolute error at the end of the primary movement was greater than absolute error at movement completion (functional); 2) absolute error at the end of the primary movement was less than absolute error at movement completion (non-functional). These data were treated as an additional repeated measure (function) in our ANOVA design, which indicated a significant difference in the overall percentage of functional (mean = 69%) vs. non-functional (mean = 31%) submovements [F(1,8) = 15.10, P < 0.01]. While there was no effect of group, there was a significant interaction between function, condition and direction [F(1,8) = 5.35, P < 0.05]. Post hoc testing indicated that both groups made a greater percentage of functional (mean = 84%) than non-functional (mean =16%) corrections when aiming toward irrespective of vision. The difference between functional and non-functional corrections when aiming toward was larger in the vision than no vision condition but both were significant. There was no difference between functional (mean = 55%) and non-functional (mean = 46%) corrections in the away condition, irrespective of vision condition.
Finally, constant error at movement completion was examined to determine the effectiveness of submovements when present. ANOVA indicated a main effect of direction only [F(1,8) = 12.340, P < 0.01], which was reflective of overshoot when aiming away from the body (grand mean = 4 mm) and undershoot when aiming toward the body (grand mean = −8 mm). There was no difference between the groups or any advantage of having vision while aiming.
Goal-directed aiming movements performed by C6 tetraplegics who have undergone tendon transfer surgery (i.e., posterior deltoid replaces the elbow extensor function of the triceps) are of longer duration and lower peak velocity than those performed by age-matched neurotypical controls. Such movement kinematics could reflect a limb control strategy of tetraplegic participants that minimizes the spatial variability associated with a noisier neuromuscular system[8, 9, 11]. However, slower aiming movement of tetraplegics compared to neurotypicals could simply be a result of adapted motor control and thus could be accompanied by a different distribution of functional and non-functional submovements[11, 12]. Here, then, to better understand the control of goal-directed aiming movements in tetraplegics with PD transfer, we report for the first time a detailed analysis of the frequency and type of submovements, as well as the movement kinematics in only those trials containing a submovement phase. These data are compared to those of a neurotypical control group, and related to other measures of motor control in order to better understand the adapted motor control of tetraplegics with PD transfer.
We then sought to determine whether trials containing submovements were associated with low movement speed. To this end, we performed individual participant analyses to determine if trials containing a submovement had higher velocity (and acceleration) than those completed with a primary movement alone. For both groups, higher speed and acceleration was not associated with the presence of submovements. We then questioned if submovements brought the hand closer to the target [see 10 for the rationale behind our method]. It was found that both tetraplegic and control groups exhibited a greater percentage of trials in which there was a decrease (functional) compared to increase (non-functional) in end-point error following a submovement. In addition, both groups made a greater percentage of functional than non-functional corrections when aiming toward in the vision and no vision conditions. There was no difference between functional and non-functional corrections in the away condition, irrespective of vision condition. These data indicate that submovements made by both groups were equally effective in the presence of vision and/or proprioception, and particularly when aiming toward the body. The similarity in end-point control between tetraplegic and control groups was confirmed by analysis of constant error at movement completion, which revealed a difference in direction only. Both groups exhibited a small overshoot when aiming away from the body and a small undershoot when aiming toward the body. Together, these data indicate that while the tetraplegic group exhibited initial impulse control (i.e., feedforward) that was more variable, this was compensated by the limb-target control phase (i.e., feedback) to achieve similar end-point accuracy as the control group; for a detailed review of feedforward and feedback processes operating in aiming see.
We acknowledge that submovements that bring the hand closer to the target could appear to be functional simply because they were identified soon after peak velocity within a trajectory that contains fluctuations (i.e., type 2 and type 3). However, with the target located at only 200 mm from the home position, it would certainly have been possible for participants to make large overshoot errors that required a type 1 (i.e., movement reversal) correction. Such behaviour was not, or at least very rarely, exhibited by either group in the current study. The virtual absence of type 1 submovements is reflective of a strategy often seen in manual aiming where the whole mass of the limb must move though 2-dimensional or 3-dimensioanl space. Specifically, performers often strategically undershoot the target with their primary movement so that time and energy consuming reversals to the direction of the movement are avoided. In the current study, participants came close to the target with the primary movement when aiming toward the body and then made type 2 or type 3 submovements that moved the hand even closer to the target (i.e., functional submovements). When aiming away from the body, participants made a larger amplitude primary movement, which coincided with fewer trials containing a submovement. It is possible that part of the direction difference reflects the temporal and energy costs associated with a movement reversal. That is, having planned to land closer to the target when aiming away, inherent movement variability resulted in some trials overshooting, which participants then chose not to correct because of the need to move the limb a greater distance, as well as overcoming the inertia associated with a zero velocity at the point of a reversal. This strategy would be consistent with the small but reliable positive constant error for away movements exhibited by participants in both groups, and demonstrates excellent spatial awareness and control of the limb on the ball transfer unit in relation to the target.
As alluded to above, the results of the current study provide some insight into feedforward and feedback motor control following the posterior-deltoid transfer. Specifically, one may have expected that feedforward control would be impaired by the transfer such that tetraplegics are unable to generate appropriately timed and scaled muscle forces, thus producing an initial impulse phase that necessitates significantly more correction during the limb-target control phase. Although we found increased spatial variability in the initial phase of reaching, reduced peak acceleration and a tendency for lower amplitude primary movement, there was no difference between the groups in the proportion of trials containing submovements. Moreover, when a submovement was made during the limb-target control phase, these were mostly functional thus indicating tetraplegic participants were able to use visual and/or proprioceptive feedback to aim at the horizontal targets with similar end-point error as neurotypical controls. We have reported a similar movement strategy in 3-dimensional aiming, thus adding support to clinical observations of a benefit for PD transfer in functional reaching tasks. As these results are from a small sample, it is not known how generalizable they are to the wider population of tetraplegics with a posterior-deltoid transfer.
In terms of functional implications, it should first be borne in mind that the inertia of the limb did not have to be independently supported in this experiment, unlike in some earlier studies of reaching[17, 18]. In this respect, the constraints of the current study are somewhat similar to those that would be experienced by the tetraplegic participants when performing tasks such as manipulating a computer mouse or motorised wheelchair control. This type of supported limb control is therefore important for a number of daily life activities including communication, education and independent transport. The results of the current study indicate that in such tasks, PD transfer enables C6 tetraplegics to effectively use their altered neurology and adapt the function played by the posterior deltoid prior to surgery. In future studies it will be important to determine the timescale for such adaptation. Additionally, more specific insight into the effect of the posterior-deltoid transfer could be gained from comparison to C7 tetraplegics who would have normal elbow extension function. This group would more closely reflect the functional capabilities of the C6 tetraplegics but without the impairment of elbow extension.
A detailed analysis of trials containing a primary and submovement phase show that although the tetraplegic group moved slower and exhibited more spatial variability during the initial phase of the movement, they achieved comparable end-point error as control participants. The implication is that tetraplegic participants adopt a subtly different but still effective speed-accuracy relation compared to control participants, indicating that feedforward and feedback motor control was both intact and effective.
The authors wish to express their thanks to Mrs. Barbara Hoole and Mr. Pradip Sett of the North West Regional Spinal Injuries Centre, Southport and Ormskirk Hospital NHS Trust for their assistance with the recruitment of participants and the Research Institute for Sport and Exercise Sciences for financially supporting this publication.
- Wierzbicka MM, Wiegner AW: Effect of weak antagonist on fast elbow flexion movements in man. Exp Brain Res 1992, 91: 509-519.View ArticlePubMedGoogle Scholar
- Hoffmann G, Laffont I, Roby-Brami A: Coordination of reaching movements in patients with a cervical spinal cord injury. Cah Psychol Cogn 2002, 21: 305-340.Google Scholar
- Laffont I, Briand E, Dizien O, Combeaud M, Bussel B, Revol M, Roby-Brami A: Kinematics of prehension and pointing movements in C6 quadriplegic patients. Spinal Cord 2000, 38: 354-362.View ArticlePubMedGoogle Scholar
- Moberg E: Surgical treatment for absent single-hand grip and elbow extension in quadriplegia - principles and preliminary experience. J Bone Joint Surg Am 1975, 57: 196-206.PubMedGoogle Scholar
- Robinson MA, Hayes SJ, Bennett SJ, Barton GJ, Elliott D: Sensory-motor equivalence: manual aiming in C6 tetraplegics following musculotendinous transfer surgery at the elbow. Exp Brain Res 2010, 206: 81-91.View ArticlePubMedGoogle Scholar
- Elliott D, Hansen S, Mendoza J, Tremblay L: Learning to optimize speed, accuracy and energy expenditure: a framework for understanding speed-accuracy relations of goal-directed aiming. J Mot Behav 2004, 36: 339-351.View ArticlePubMedGoogle Scholar
- Elliott D, Hansen S, Grierson L, Lyons J, Bennett SJ, Hayes SJ: Goal-directed aiming: Two components but multiple processes. Psychol Bull 2010, 136: 1023-1044.View ArticlePubMedGoogle Scholar
- Schmidt RA, Zelaznik HN, Hawkins B, Frank JS, Quinn JT: Motor output variability: A theory for the accuracy of rapid motor acts. Psychol Rev 1979, 86: 415-451.View ArticleGoogle Scholar
- Worringham C: Variability effects on the internal structure of rapid aiming movements. J Mot Behav 1991, 23: 75-85.View ArticlePubMedGoogle Scholar
- Khan MA, Franks IM, Elliott D, Lawrence GP, Chua R, Bernier P-M, Hansen S, Weeks DJ: Inferring online and offline processing of visual feedback in target-directed movements from kinematic data. Neurosci Biobehav Rev 2006, 30: 1106-1121.View ArticlePubMedGoogle Scholar
- Meyer DE, Abrams RA, Kornblum S, Wright CE, Smith JEK: Optimality in human motor performance: ideal control for rapid aimed movements. Psychol Rev 1988, 95: 340-370.View ArticlePubMedGoogle Scholar
- Dounskaia N, Fradet L, Lee G, Leis BC, Adler CH: Submovements during pointing movements in Parkinson’s disease. Exp Brain Res 2009, 193: 529-544.View ArticlePubMedGoogle Scholar
- Welsh TN, Higgins L, Ray M, Weeks D: Seeing vs. believing: Is believing sufficient to activate the processes of response co-representation? Hum Mov Sci 2007, 26: 853-866.View ArticlePubMedGoogle Scholar
- Fradet L, Lee G, Dounskaia N: Origins of submovements in movements of elderly adults. J Neuroeng Rehabil 2008, 5: 28.View ArticlePubMedPubMed CentralGoogle Scholar
- Fradet L, Lee G, Dounskaia N: Origins of submovements during pointing movements. Acta Psychol (Amst) 2008, 129: 91-100.View ArticleGoogle Scholar
- Chua R, Elliott D: Visual regulation of manual aiming. Hum Mov Sci 1993, 12: 365-401.View ArticleGoogle Scholar
- Robinson MA, Barton GJ, Lees A: Sett, P: Analysis of tetraplegic reaching in their 3D workspace following posterior deltoid-triceps tendon transfer. Spinal Cord 2010, 48: 619-627.View ArticlePubMedGoogle Scholar
- Remy-Neris O, Milcamps J, Chikhi-Keromest R, Thevenon A, Bouttens D, Bouilland S: Improved kinematics of unrestrained arm raising in C5-C6 tetraplegic subjects after deltoid-to-triceps transfer. Spinal Cord 2003, 41: 435-445.View ArticlePubMedGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly credited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.