The feedforward control

of

posture and movement

Julia Anne Schaefer Leonard

Department of Kinesiology and Physical Education

Faculty of Education

McGill University, Montreal

August 2012

A thesis submitted to the faculty of Graduate Studies and Research

In partial fulfilment of the degree of

Doctor of Philosophy

© Julia Leonard 2012. All rights reserved

i

Table of Contents

LIST OF FIGURES ....................................................................................... v

LIST OF TABLES ........................................................................................ ix

Abstract ......................................................................................................... xi

Résumé ........................................................................................................ xiii

Statement of originality ............................................................................... xv

Acknowledgements .................................................................................... xvii

Contributions of authors ............................................................................ xix

List of symbols and abbreviations .............................................................. xxi

Chapter 1 ........................................................................................................ 1 1.1 Scientific Background ..................................................................................2 1.2 Rational ........................................................................................................5 1.3 General Aim .................................................................................................5 1.4 Scientific Objectives and Hypotheses ..........................................................6

Chapter 2 ........................................................................................................ 9 2.1 How are voluntary movements and posture controlled? ......................... 10

2.1.1 The neuroanatomical basis of movement execution .................................. 10 2.1.2 Circuitry of the spinal cord provides a basis for coordinating movement ... 12 2.1.3 Somatotopic organization of spinal cord ................................................... 13 2.1.4 Anatomical organization of the descending pathways for the control of movement ......................................................................................................... 15 2.1.5 Integration of central commands for the global planning of movement and

posture .............................................................................................................. 19 2.2 Postural Control ........................................................................................ 25

2.2.1 Biomechanical requirements for equilibrium control................................. 25 2.2.2 Behavioural goals of the postural system .................................................. 26 2.2.3 Sensorimotor control of posture ................................................................ 27 2.2.4 The problem of motor redundancy ............................................................ 30

2.3 Mechanisms of postural control ................................................................ 32 2.3.1 Overview ................................................................................................. 32 2.3.2 Intrinsic mechanical properties for stability .............................................. 33 2.3.3 Feedback postural responses ..................................................................... 34 2.3.4 Feedforward postural adjustments ............................................................ 37

2.4 Models for movement control ................................................................... 41 2.4.1 Goal-directed movements require both feedback and feedforward control

mechanisms ...................................................................................................... 42 2.4.2 Internal models ........................................................................................ 43

2.5 The control of voluntary arm movements ................................................ 45 2.5.1 Online control of visually-guided reaching movements ............................. 46 2.5.2 Standing imposes equilibrium constraints during perturbed reaching ........ 47

2.6 Summary and direction for future investigation ...................................... 48

ii

Chapter 3 ...................................................................................................... 51 3.1 Rational for experimental protocol ........................................................... 51 3.2 Overview of experimental protocol ........................................................... 53

3.2.1 Experimental apparatus ............................................................................ 53 3.2.2 Behavioural task....................................................................................... 55 3.2.3 Protocol specific to SA1 and SA2 ............................................................. 55 3.2.4 Protocol specific to SA3 ........................................................................... 56 3.2.5 Data collection and analysis ..................................................................... 57

3.3 Significance of the experimental paradigm provides basis for further

exploration .......................................................................................................... 57

Chapter 4 ...................................................................................................... 59 4.1 PREFACE ..................................................................................................... 59 4.2 ABSTRACT................................................................................................... 60 4.3 INTRODUCTION ......................................................................................... 60 4.4 MATERIALS AND METHODS .................................................................. 63

4.4.1 Subjects ................................................................................................... 63 4.4.2 Experimental apparatus and set-up ........................................................... 63 4.4.3 Experimental Procedures .......................................................................... 65 4.4.4 Data analysis ............................................................................................ 68

4.4.5 Statistical analysis .................................................................................... 69

4.5 RESULTS ...................................................................................................... 70 4.5.1 Kinematics of reaching movements during standing ................................. 70 4.5.2 EMG activity in relation to the forces produced: pPA period .................... 74 4.5.3 EMG activity in relation to the forces produced: aPA period .................... 76 4.5.4 Feedforward postural adjustments show directional tuning and are synergic

......................................................................................................................... 76 4.5.5 Spatial patterns of force differ between preparatory and associated postural

adjustments ....................................................................................................... 82 4.6 DISCUSSION ................................................................................................ 88

4.6.1 The roles of preparatory and associated postural adjustments for reaching

during stance .................................................................................................... 88 4.6.2 Tuned, synergic muscle activity characterizes feedforward postural

adjustments ....................................................................................................... 89 4.6.3 Clearly constrained force patterns are seen during preparatory but not

during associated feed- forward postural adjustments ........................................ 90 4.6.4 Implications for the neural control of balance: shared control of feedforward

and feedback postural adjustments .................................................................... 93 4.7 ACKNOWLEDGEMENTS .......................................................................... 94

Chapter 5 ...................................................................................................... 95 5.1 PREFACE ................................................................................................. 95 5.2 ABSTRACT ............................................................................................... 95 5.3 INTRODUCTION ..................................................................................... 96 5.4 METHODS ................................................................................................ 99

5.4.1 Subjects ................................................................................................... 99 5.4.2 Experimental apparatus and set up ............................................................ 99 5.4.3 Data processing and analysis .................................................................. 100

5.5 RESULTS ................................................................................................ 105 5.5.1 Feedforward postural muscle activity is directionally tuned, but shows

variability between trials ................................................................................. 105 5.5.2 Composition and tuning of muscle synergies .......................................... 108

iii

5.5.3 Muscle synergies accurately predict muscle activity patterns .................. 113 5.5.4 Comparison of muscle synergy structure between subjects ..................... 116

5.6 DISCUSSION .......................................................................................... 118 5.6.1 Modular organization of feedforward postural adjustments ..................... 118 5.6.2 Similar organization for feedforward and feedback postural control ........ 120 5.6.3 Conclusions ........................................................................................... 122

5.7 ACKNOWLEDGEMENTS .................................................................... 123

Chapter 6 .................................................................................................... 125 6.1 PREFACE ............................................................................................... 125 6.2 ABSTRACT ............................................................................................. 126 6.3 INTRODUCTION ................................................................................... 127 6.4 METHODS .............................................................................................. 129

6.4.1 Subjects ................................................................................................. 129 6.4.2 Experimental apparatus and set up .......................................................... 129 6.4.3 Experimental procedures ........................................................................ 131 6.4.4 Data analysis .......................................................................................... 135 6.4.5 Statistical analysis .................................................................................. 138

6.5 RESULTS ................................................................................................ 139 6.5.1 Unperturbed reaching and characteristics of online corrections ............... 139 6.5.2 Corrective forces and electromyographic activity accompanying online

corrections of arm movements ........................................................................ 144 6.5.3 Arm-muscle activity responsible for corrections of finger trajectory ....... 147 6.5.4 Corrective postural adjustments in leg muscles lead arm muscle corrections

during online corrections of arm trajectory to unexpected shifts of target position

....................................................................................................................... 147 6.6 DISCUSSION .......................................................................................... 153

6.6.1 Methodological considerations ............................................................... 154 6.6.2 Postural adjustments contribute to the execution of voluntary movement 155 6.6.3 Effects of standing on the characteristics of online corrections of the arm

....................................................................................................................... 156 6.6.4 Implications for the control of posture and movement ............................ 158 6.6.5 Conclusions ........................................................................................... 162

6.7 ACKNOWLEDGEMENTS .................................................................... 162

Chapter 7 .................................................................................................... 163 7.1 Characterization of feedforward postural adjustments during multi-

directional reaching movements ....................................................................... 164 7.1.1 Role feedforward postural activity .......................................................... 164 7.1.2 Independent or parallel commands for global planning of posture and movement ....................................................................................................... 166 7.1.3 Strategies for simplifying the control of posture and movement .............. 167 7.1.4 Force constraint strategy: neural strategy or geometry? ........................... 168 7.1.5 The importance of muscle tuning and synergic organization for feedforward

postural control ............................................................................................... 170 7.2 Central control of posture and movement: integration of feedback and

feedforward postural commands ...................................................................... 173 7.3 Predictive motor control: internal model of posture............................... 174 7.4 Justification for understanding disorders of posture and balance ......... 176 7.5 Conclusions and future directions ........................................................... 177

7.5.1 Do the elderly differ in the spatial and temporal organization of feedforward

postural control? ............................................................................................. 177

iv

7.5.2 Are feedforward muscle synergies robust and how does their recruitment

relate to task-level goals? ................................................................................ 179 7.5.3 Online control of posture: effects of direction and time of visual

perturbation .................................................................................................... 180

Chapter 8: References ............................................................................... 183

v

LIST OF FIGURES

Figure 3. 1: General experimental set-up. Plan view of target light array illustrating subject

orientation on the force plate relative to the array. Targets are arranged from right (0°) to

left (180°). ................................................................................................................... 54

Figure 4. 1: Plan view of the target array and temporal sequence of data collection. A. Subjects

stood on 2 force plates, 1 under each foot and were centered in a 180° light target array,

adjustable for each subject in height and distance (see Methods). Targets (light emitting

diodes, LEDs) were placed at 15o intervals from right to left sides with the position of

each LED set to exactly 130% of their outstretched arm length at shoulder height. Fy =

anterioposterior force, Fx = mediolateral force and Fz = vertical force. B. Temporal

sequence of the data collection period. An auditory tone 500 ms in length sounded to

inform subjects of an impending target illumination. A period of 1000 ms preceded the

onset of the target light upon which subjects were required to reach and point to the

target. The total acquisition period was 3000 ms. A representation of an approximate

movement length (movement time, MT) is shown. ....................................................... 67

Figure 4. 2: 3D kinematic representations of reach to point movements to 3 principal target

directions. (A. 0°, B. 90° and C. 180°) for 1 subject (S5). Stick figures are shown as if

being viewed from the front and slightly to the subject’s left (A) and right (B and C)

sides. Body movements are shown at 2 different times: At the onset of finger movement

(grey sticks) and at the end of the movement (black sticks) when the finger touched the

target. Shown are the following segments bilaterally: foot, shank, thigh, shoulder, upper

arm and lower arm. Markers at the level of the 7th cervical and 10th thoracic vertebrae, the clavicle and sternum form a single segment that represents the torso in the sagittal plane.

The head is represented by markers placed at 4 locations on the left and right temples and

at the same level at the back of the head. Finger trajectory is shown in red. The body

centre of mass (CoM) is shown as a grey and black circles (onset and end of the

movements, respectively). The ground reaction force vector is represented at the onset of

movement in red, and at target attainment in blue. D. Trajectories of the CoM from the

onset of the focal movement (open squares) to when the finger touched the target (open

circles). Trajectories are represented for 5 trials in each direction and are colour-coded as

per the legend across the bottom of the figure. .............................................................. 73

Figure 4. 3 Electromyographic activity, changes in force and vertical torque (Tz) under each

foot for reaching movements to 3 principal directions (A. 0°, B. 90° and C. 180°). Traces

are shown for a period of 500 ms preceding movement onset until the end of each

movement for one typical trial in subject 5. On each plot, the full grey vertical line

indicates the onset of the light target (Light on). The dashed grey line to the left of

movement onset (Movt on) indicates the onset of force and Tz changes during the pPA

period. The dashed grey line to the right of movement onset (between 500 and 750 ms)

indicates the end of the arm movement (Movt end). Forces are shown as forces exerted

against the ground. TFLr and TFLl = tensor facia latae muscles (right and left, respectively), RFr and RFl = rectus femoris, BFr and BFl = biceps femoris, GasLr and

GasLl = gastrocnemius lateralis, , Perr and Perl = peroneus longus, TAr and TAl = tibialis

anterior, Solr and Soll = soleus. Fx = mediolateral force, Fy = anterioposterior force, Fz =

vertical force and Tz = torque exerted around the vertical axis. Left =leftwards, back =

backwards, load = loading and CCW = counterclockwise. Left and right foot forces are

represented by solid and dashed traces, respectively (see legend). Shaded regions

vi

represent the 2 periods of 250 ms under study (pPA and aPA). Successive squares

underneath the time axes pictorially represent changes in Tz at each foot during the two

periods. ....................................................................................................................... 75

Figure 4. 4 Representative EMG traces for 14 selected muscles for S5 across the 13 directions

of pointing. Muscle activity is shown for a total duration of 500 ms, 250 ms before and

after the onset of the pointing movement. Muscle name conventions are as described in

Figure 4.3. The shaded area to the left of time zero on each muscle plot represents the 250

ms preparatory period. Unless shown, muscles have the same scaling for the left leg (top

row) as they do for the right leg (bottom row). ............................................................. 78

Figure 4. 5 Muscle tuning curves for the EMG activity of all 14 postural muscles during the

final 3 bins of the preparatory and associated periods for the representative subject S5.

Differences in tuning and recruitment of the muscle studied can be observed by

comparing the activity of the muscles over the 3 5 equivalent bins (left to right columns).

Dots indicate amplitudes from each trial measured and the solid lines the mean responses.

Muscle name conventions are as described in Figure 4.3............................................... 80

Figure 4. 6 Muscle tuning curves in the final bin of the pPA (A) and aPA (B) periods for each

of the 8 subjects studied. Muscles have been grouped into the 3 major groups that

activated for similar directions of reach. Tuning curves and individual trials are

represented as in Figure 4.5. Schema (C) summarizing the approximate range of

directions of reach to which each identified group contributed. ..................................... 81 Figure 4. 7 Individual resultant horizontal ground reaction force vectors and average values of

Fz produced during the pPA period (A, B, respectively) and the aPA period (C, D) for

subject 5. Forces are shown for each consecutive bin during each period in successive

rows from top to bottom. Black and grey arrows represent the approximate direction of

exerted force under the loaded and unloaded feet, respectively. In B and D, bars above the

top of the plots marked ‘L’ indicate directions of reach for which Fz under each

respective foot was loaded. For reference, the directions of reach used are indicated on the first plot (left foot) for the pPA1 period in Fig. 4.6A. .................................................... 84

Figure 4. 8 Average direction and magnitude of horizontal ground reaction force change during

each bin of the pPA and aPA periods under both feet. .................................................. 86

Figure 5. 1 Representative traces of muscle activity in seven muscles recorded bilaterally

during a 500 ms period preceding movement onset (Movt On) to movement end. The

pPA period is indicated by the shaded vertical grey area to the left of movement onset.106

Figure 5. 2 Muscle tuning curves for all 14 muscles and all 5 time bins during the 250 ms

period preceding the onset of reaching movements in a representative subject (S011). .107

Figure 5. 3 Variability accounted for (VAF) for different number of muscle synergies for the

entire data set for a representative subject (S010). To choose an appropriate number of

muscle synergies (Nsyn), the following criteria were met: (A) Overall VAF attained a

threshold of > 90%; (B) VAF by muscle as a function of muscle synergy number. (C)

VAF by synergy number as a function of muscle was plotted to confirm that chosen Nsyn

met the criteria of 75%. (D) VAF by direction and bin shows that directions were well

characterized. .............................................................................................................109

Figure 5. 4 A: Muscle synergy vectors (W) and B: recruitment coefficients (C) for a

representative subject (S010) across all 5 time bins of the pPA. Synergy activation

coefficients for individual trials are shown by a dot, and average muscle synergy

recruitment is shown by a solid line that illustrates its directional tuning. .....................112

vii

Figure 5. 5 Reconstruction of mean muscle tuning curves using the muscle synergies shown in

Fig. 5.4 for a representative subject (S010). Dashed lines represents observed data and

solid lines represent reconstructed data. Each muscle synergy’s contribution is shown by

the corresponding colored line. Combined, these result in the total reconstruction.

Goodness of fit (VAF, r2) of the reconstruction to the observed EMG is indicated. ......114

Figure 5. 6: Postural muscle activity in the REM trials for reaching to the 75° target is

reconstructed using the extracted muscle synergies for a representative subject (S010).

Variations in the activation levels of the muscles between trials is well accounted for by

the modulation of the muscle synergies. Muscle activation amplitudes for all recorded

muscle are grouped along the x-axis within a trial. Star (✴) represents the observed EMG

and open circle ( ○) is the reconstructed activation amplitude. The relative contributions

of the muscle synergies to the overall muscle activity is shown by the colours in the

vertical histogram for each muscle. .............................................................................115

Figure 5. 7 Muscle synergy structure compared between subjects. Muscle synergies that are

shared between subject are indicated by a significant VAF and r2. Muscle synergies

whose backgrounds are shaded gray are specific to that subject. ..................................117

Figure 6. 1: Experimental set-up and data collection schema. A. Subjects stood on 2 force

plates reached to a central target, aligned with their xiphoid process. Unperturbed ‘reach’

trials were interspersed with online correction (‘corr’) trials involving unexpected illuminations of 1 of 3 other targets placed successively at 15° increments to the right of

centre. B. Explanation of the changes in voltage related to the sequence of light changes.

When the signal rose to 5V each light was illuminated. L1 = light one, L2 =light 2, chest

= chest switch attached around the subjects sternum that acted as a signal from which L2

illumination could be triggered. C. A histogram showing the distribution of L2 onset as a

percentage of mean ‘reach’ peak velocity. Trials from all ‘corr’ conditions have been

pooled (n=652). rFin = right finger. .............................................................................133

Figure 6. 2: Determination of the online correction of finger trajectory (fcorrect). A. Plan view

(x,y) of rFin average ‘reach’ trajectories +/- 1SD (dashed line with shaded grey area) in

relation to one ‘corr45’ trial (full black trajectory). Filled black circle is the onset of light

2 (L2 onset) and the open circle is the time at which the corr45 x,y trajectory exceeded

the average ‘reach’ trajectory plus 1SD for subject S6. B. Average (dashed line) plus 1SD

of curvilinear rFin velocity for a reach movement and one ‘corr45’ trial(full black line).

Black vertical line is light 2 (L2) onset, grey vertical line is the time of online correction

(fcorrect). Each corr condition has been displaced rightwards and downwards for clarity,

but the starting position was the same for each. C, D: Explanation of how the correction

of the EMG activity and GRFcorrect associated with online corrections were determined.

C. Calculation of EMGcorrect. The muscle shown is the left soleus muscle, but the same

procedure was used with all other muscles studied (see Methods). The dashed trace and grey traces represent respectively, the mean ‘reach’ soleus muscle activity ± 2SDs above

and below the mean. The dark full trace represents the soleus muscle activity produced

during an online correction movement, in this example a corr45 movement. The open

circle indicates the time at which the corr45 soleus muscle activity exceeded the

mean+2SD ‘reach’ activity level. This time was taken as the EMGcorrect time (for more

detailed explanation, see Methods and Results). Abbreviations as previous figures, except

Movt end=movement time. D. Calculation of GRFcorrect. Method for determining

GRFcorrect is shown for the left shear force (Fx). The dashed trace and grey traces

represent respectively the mean ‘reach’ Fx and ± 2SDs above and below the mean. The

dark full trace represents the Fx exerted during an online correction movement (in this

example corr45). The open circle indicates the time at which the exerted force was significantly different from the mean forces exerted in a ‘reach’ trial. ..........................136

viii

Figure 6. 3: Reaching movement kinematic characteristics and profiles of curvilinear velocity.

Shown are averages plus 1SD for all trials for subject 2 in each of the 4 conditions

studied. A. Plan view (x,y) kinematics of rFin trajectory for ‘reach’ trials and each of the

correction conditions. B. rFin curvlinear velocity also for all 4 conditions.

fcorrect=kinematic correction of finger trajectory, mvt end=end of the focal movement

(reach and corrected movements). ...............................................................................141

Figure 6. 4 Typical arm and leg muscle activity in relation to the 3D ground reaction forces

produced for a ‘reach’ movement (A) and an online correction movement to the target

placed 45° to the right of midline, ‘corr45’ (B). In each, the muscles plotted in grey are

those recorded in the right leg. The vertical dashed line indicates initial movement onset

(‘Mvt Onset’) and the full vertical black line, movement end (‘Mvt End’). In B., the

shaded grey area indicates the area in which arm and postural adjustments occurred. For

muscle abbreviations, see Methods. L2 onset=light 2 onset, fcorrect=time of kinematic correction of the rFin marker. Fx=mediolateral force, Fy=anteriorposterior force and

Fz=vertical force. ........................................................................................................146

Figure 6. 5: Linear regressions calculated between the four arm muscles recorded in the right

am and the fcorrect latencies calculated using the curvilinear kinematics of the rFin

marker. A. right posteior deltoid, B. right triceps, C. right anterior deltoid and D. right

biceps. Yi=the value of the Y intercept when X is zero. ...............................................149

Figure 6. 6: Multiple comparisons of differences between EMGcorrect values of arm and leg

muscles. A. Average differences EMGcorrect differences per ‘corr’ target. Values for all

3 arm and leg muscles have been pooled (averages for each ‘corr’ target +/- 95%

confidence interval, CI). Positive differences indicate postural muscle corrections before

arm corrections (see direction of arrow, top right of figure). B. Average differences (+/-

810 .................... 95% CI) EMGcorrect (all arm muscles pooled) per leg muscle. Values for

EMGcorrect measures were pooled for all arm muscles and expressed as differences with

each leg muscle in turn (positive differences also indicate postural muscle corrections before arm muscles). C. EMGcorrect differences (+/- 95% CI) per phase of reach, i.e.

before peak velocity (Acceleration) or after peak velocity (Deceleration). Filled circles

show mean EMGcorrect differences for all targets, while open circles show only data for

corr45; shaded area represents 95% confidence interval. D. Average differences (+/- 95%

CI) EMGcorrect (all arm muscles pooled and all leg muscles pooled) per subject.........150

ix

LIST OF TABLES

Table 4. 1 Mean (±1SD) of movement times for pointing movements in all 13 directions ...... 72

Table 4. 2 Results of the broken-axis approach (Holmquist & Sandberg, 1991) ..................... 87

Table 6. 1 Breakdown of total number of trials collected and retained after trial selection

procedure. ..................................................................................................................143

Table 6. 2 Slope (m), Y intercept (Yi), r2 values, p-value for the strength of the regression fit

(p) and p-value for the intercept (p-int) for linear regressions conducted between the leg

and arm muscles selected to characterize the online corrections to all targets for each subject ........................................................................................................................152

x

xi

Abstract

Goal-directed arm movements performed in the standing position

potentially disturb the body's equilibrium as a result of the multi-linked

structure of the musculoskeletal system. To compensate for these disturbances

and ensure that stability is maintained, the central nervous system (CNS)

organizes postural adjustments preceding and accompanying the voluntary

movement in a feedforward manner (Massion 1992) using knowledge of the

dynamics of the body (Bouisset and Zattara 1981). To date, most studies

investigating the control of posture during voluntary movements in humans

have focused on either the role of the postural activity preceding the movement

or on the temporal structure of these anticipatory postural adjustments (APAs)

with respect to the focal movement. As such, detailed knowledge about the

spatial organization of postural activity is lacking. Furthermore, it is not clear

how posture is coordinated when the goal of a voluntary movement changes

online. Therefore, the studies in this thesis were aimed at addressing these

questions to develop a greater understanding of the organization of feedforward

postural control during voluntary movements.

Muscle activity, kinetics and kinematics were recorded as subjects

performed unperturbed and perturbed reaching movements to targets located in

multiple directions while standing. Feedforward postural control strategies

preceding and accompanying the reaching movements were quantified.

Characterization of the spatial and temporal patterns of muscle activity and

ground reaction forces of postural adjustments preceding reach movements

revealed that muscle activity was directionally-tuned to reach direction and

forces that were constrained to two principal directions. Also, muscle synergies

were able to explain the spatial and temporal variability in postural muscle

activity in the period preceding the reaching movements, suggesting that a

modular organization of muscle recruitment is adopted for this task. Overall,

these strategies are similar to those observed for feedback postural responses,

xii

suggesting that the CNS relies on shared neural structures for controlling

posture in both modes of control. Lastly, the nature of postural control was

examined when reaching movements were perturbed with a shift of the visual

target after the reaching movement was initiated. Here, muscle activity in the

legs was consistently modulated prior to changes in the muscle activity related

to the online correction of the arm trajectory.

Taken together, the findings of this thesis provide important insights

into how the brain coordinates the control of posture and movement. This work

provides a measure of feedforward postural control strategies in healthy, young

adults as a first step to understanding how and why deficits in balance control

may occur during the execution of voluntary movements in fall-prone

individuals.

xiii

Résumé

Les mouvements volontaires effectués dans la position debout peuvent

engendrer des perturbations de l’équilibre en raison de la structure complexe du

système musculo-squelettique. Pour amorcer ces perturbations et s’assurer que

l’équilibre est maintenu, le système nerveux central (SNC) amorce le

déplacement du centre de masse (CM) par la mise en jeu d’ajustements

posturaux avant et accompagnant les mouvements programmés en mode

proactif (Massion 1992) en utilisant des représentations internes du corps et de

l’environnement. À ce jour, la majorité des études portant sur le contrôle de la

posture lors des mouvements volontaires chez l’homme ont comme but soit

l’identification du rôle ou la caractérisation de la structure temporelle de ces

ajustements posturaux anticipateurs. Cependant, une connaissance approfondie

concernant l’organisation spatiale de l’activité posturale est manquante. De

plus, ce n’est pas évident comment la posture est coordonnée lorsque le but du

mouvement change après le commencement du mouvement. Ainsi, les études

présentées ici ont comme but de répondre à ces questions pour développer une

meilleure compréhension de l’organisation centrale de la posture et le

mouvement.

Les signaux électromyographiques, les forces de réaction au sol et la

cinématique tridimensionnelle ont été enregistrés pendant que les sujets

effectuaient des mouvements de pointage vers des cibles distinctes dans la

position debout. Les stratégies posturales organisées en mode proactif ont été

quantifiées sans pertubations et avect des pertubations visuomotrices des

movements d’atteinte. La caractérisation de l’organisation spatiale et temporelle

de l’éléctromyographie et des forces appliquées au sol ont révélé que l’activité

des muscles était biaisée vers la direction de pointage (‘directionally-tuned’)

mais que les forces au sol étaient appliquées dans un nombre de directions

limitées (‘force constraint strategy’). De plus, la variabilité spatiale et

temporelle de l’activité des muscles posturaux était expliquée par les synergies

xiv

musculaires. Ceci suggère qu’une organisation modulaire est utilisée par le

SNC pour faciliter la tâche de contrôle de la posture. Ces stratégies sont

similaires à celles observées pour les ajustements posturaux compensatoires (à

base de ‘feedback’ ou rétroaction), ce qui suggère que le SNC dépend des

mêmes structures neuronales pour contrôler la posture dans la mode proactif et

rétroactif. Par la suite, la nature du signal pour le contrôle de la posture a été

examinée lors des mouvements de pointage qui ont été perturbés avec un

déplacement de la cible visuelle après que le mouvement ait été commencé. Ici,

l’activité musculaire dans les jambes était modulée avant la modulation de

l’activité musculaire liée à la correction de la trajectoire du bras.

Ensemble, les conclusions de cette thèse fournissent un aperçu

important sur la façon dont le cerveau coordonne le contrôle de la posture et du

mouvement. Les résultats présentés supportent la conclusion que les

commandes centrales pour la posture et le mouvement interagissent dans le

SNC, et que les structures neuronales sont partagées pour la posture organisée

de façon anticipatoire, ou proactif, et compensatoire. Les stratégies posturales

typiques dans les jeunes adultes en santé sont quantifiées et forment une base de

données pour la comparaison avec des gens sujets au déséquilibre lors de la

performance des mouvements volontaires.

xv

Statement of originality

This thesis incorporates the outcome of research undertaken under the

supervision of Dr. Stapley in the department of Kinesiology and Physical

Education, at McGill University for the requirements of Doctor of Philosophy. I

certify that this thesis, and the research to which it refers are the product of my

own work and has not been published elsewhere except where specific

references are indicated. The manuscripts presented in chapters 4, 5 and 6

represent original material and contribute to the advancement of knowledge in

the fields of posture and movement control. To my knowledge, the studies

presented within this thesis are the first to investigate the organization of feed-

forward postural adjustments during multidirectional reaching performed in

standing in human subjects.

All data presented in this thesis were collected in the BVML (Balance

and Voluntary Movement Laboratory), located in the department of

Kinesiology and Physical Education at McGill University. The protocols used

in the studies herein were approved by the McGill University Research Ethics

Board.

xvi

xvii

Acknowledgements

Over the course of my studies at McGill, I have had the opportunity to

interact with and learn from many talented and inspiring individuals. I am

deeply thankful for all of these experiences, as they have been influential in

shaping my ideas about science, my ability to ask questions and test reasonable

hypotheses.

In particular, I would like to thank my thesis supervisor, Dr. Paul

Stapley. It goes without saying that the work presented in this thesis would not

have been possible without his unrelenting support and patience. His guidance

has been instrumental in my development as a scientist from a naive first-year

Master’s student. In particular, he encouraged independence while maintaining

a level of support I could always count on. I will be forever grateful for the

skills I developed while working in his laboratory. He helped me create a

foundation that I hope to build upon as I continue to mature as a scientist.

I would like to thank my committee members: Drs. Trevor Drew, David

Ostry and Ted Milner for their insight and feedback. In particular, I wish to

thank Dr. Trevor Drew, who provided important feedback on the manuscripts

published in this thesis. I am also very thankful for the extensive support Dr.

Drew has provided as I transition beyond my doctoral studies. I admire his

passion for scientific inquiry and I look forward to his guidance and

mentorship. I wish to also thank Dr. Milner for maintaining a space where I

could work in the final stages of the preparation of this thesis.

I am deeply grateful for the assistance of many people in the

Department of Kinesiology and Physical Education. In particular, Ryan

Ouckama, provided invaluable technical knowledge and accessibility for

trouble-shooting for the experimental set-up; J.J. Loh, shared his love of Matlab

and taught me foundations in computer programming; past and present

members of the BVML lab, including Ryan Brown, Alicia Hilderley, Silvia

xviii

Hua and Will Lee-Shanok, thank you for sharing such an important part of my

intellectual and personal development.

And to the many others, friends and collaborators beyond BVML, thank

you. Dr. Lena Ting, for graciously welcoming me to her lab to learn the

techniques for the synergy analysis; ‘Team Synergy’ - in particular, Drs. Stacie

Chvatal and Seyed Safavynia, thank you for your prompt and thorough

responses to my many questions. I would like to thank Dr. Rob Kearney for his

insight on data analysis and Dr. Jane Macpherson for her comments on the

studies presented in this thesis. And on a personal level, thank you Dr. Karen

Lomond, Dr. Catherine Sabiston and Marilee Nugent for all your support along

this tremendous journey. Also, this work would not have been possible without

the financial support from Fonds de la Recherche en Santé du Québec (FRSQ),

National Science and Engineering Research Council (NSERC) and Canada

Foundation for Innovation (CFI; infrastructure).

Finally, this thesis is the outcome of hard work and dedication that

would not have been possible without the unrelenting support and

unconditional love from my family and close friends. In so many ways, this

thesis reflects the efforts and contributions from so many. Firstly, thank you

Eric. You have been steadfast in your words of affirmation and belief in me.

Thank you for all the sacrifices you have made so that I may pursue this goal.

You are an amazing father and I am truly blessed to share this life with you.

Mom and Dad, you modeled perseverance and commitment to a task. I would

not be where I am today without the generous and unwavering support you

provided. This thesis, and so much more, could not have been completed

without you. To my siblings, Erica, Andrew and David, thank you for your

constant encouragement. Sue and Dave, thank you for the countless hours of

babysitting and belief in me. And lastly, thank you to my two beautiful girls,

Theresa and Madeline. You inspire me to work hard, to focus and be a good

example for you. I have already learnt so much from you both; I love you

dearly.

Julia

xix

Contributions of authors

The work presented in this thesis is primarily the work of the author,

Julia Leonard, including the conception of ideas, development of the

experimental protocol, data collection, data analysis, presentation of results and

preliminary drafts of manuscripts for publication. The outcome of this research

has resulted in two published manuscripts (Chapters 4 and 6) and provides the

basis for a third one in preparation for publication (Chapter 5).

Chapter 4 contains a published manuscript: Leonard JA, Brown RH,

Stapley PJ. Reaching to multiple targets when standing: the spatial organization

of feedforward postural adjustments. J Neurophysiol 101: 2120–2133, 2009.

For this study, I built the experimental set-up, collected and analyzed the data,

prepared figures for presenting the results and prepared and revised the

manuscript following peer-review. Ryan Brown is listed as second author for

his contributions in building the experimental set-up and initial data collection

and preliminary analyses. Contributions from JJ Loh for technical assistance

and Drs. Jane Macpherson and Trevor Drew for insightful discussions are noted

in the acknowledgments section of the manuscript. As my thesis supervisor, Dr.

Stapley provided guidance and insight in all stages of the study and preparation

of the final manuscript.

Chapter 5 contains a manuscript in preparation for the Journal of

Neurophysiology: Leonard JA, Chvatal S, Ting LH, Stapley PJ. Muscle

synergy characterization of feed-forward postural adjustments during reaching

in standing humans. For this study, I built the experimental set-up, collected

and analyzed the data, prepared figures for presenting the results and prepared

the manuscript for submission to peer-review. Dr. Stacie Chvatal provided the

basic Matlab NNMF algorithms, which I customized to my analysis, as well as

important advice about the data analysis and interpretation of results. Dr. Lena

Ting provided advice about interpretation of results and collaborated in

preparing the final version of the manuscript. Dr. Safavynia provided additional

xx

advice about the data analysis and is mentioned in the acknowledgements

section of the manuscript. As my supervisor, Dr. Stapley provided guidance and

insight in all stages of the study and preparation of the final manuscript.

Chapter 6 contains a published manuscript: Leonard JA, Gritsenko V,

Ouckama R, Stapley PJ. Postural adjustments for online corrections of arm

movements in standing humans. J Neurophysiol 105: 2375–2388, 2011. For this

study, I built the experimental set-up, collected and analyzed the data, prepared

figures for presenting the results and prepared and revised the manuscript

following peer-review. Dr. Gritsenko assisted with portions of the data analysis

and collaborated in preparing the final version of the manuscript. Ryan

Ouckama provided technical expertise in programming the experimental set-up.

Dr. Rob Kearney provided advice about the data analysis and is mentioned in

the acknowledgements section of the manuscript. As my thesis supervisor, Dr.

Stapley provided guidance and insight in all stages of the study and preparation

of the final manuscript.

xxi

List of symbols and abbreviations

APA Anticipatory postural adjustment

aPA Associated postural adjustments

APR Automatic postural response

BoS Base of support

CM Centre de masse

CNS Central nervous system

CoM Center of mass

CoP Center of pressure

CPG Central pattern generator

CST Corticospinal tract

DoF Degree of freedom

Fx Mediolateral force

Fy Anteroposterior force

Fz Vertical force

EMG Electromyography

GRF Ground reaction force

ICA Independent component analysis

LED Light emitting diode

NNMF Non-negative matrix factorization

PCA Principal component analysis

pPA Preparatory postural adjustment

PMRF Pontomedullary reticular formation

SNC Système nerveux central

STA Spike-triggered averaging

Tz Moment of force about the vertical (z) axis

xxii

1

Chapter 1

General Introduction

The excitement felt by a parent watching their infant stand for the first

time, and the sense of accomplishment radiating from the infant’s face, suggest

that there is an inherent understanding that the seemingly simple act of

standing, or balancing on two feet, is in fact, a major accomplishment. Armed

with the skill of upright stance, the infant enters an entirely new world of

discovery and exploration, using their hands to grasp, manipulate and move

objects while standing. Stance control, as developed from infancy, is a

fundamental skill that affords the ability to interact with our environment in

very intimate, creative and useful ways. Moreover, while most people take the

skill of reaching while standing for granted, an aging person afflicted with

balance deficits will tell you that performing such a dynamic task is challenging

and poses significant threats to their stability. This is particularly apparent in

situations where the task may change once a movement has already been

initiated, such as reaching for a support that may change positions after the

reach began. In fact, falls, which have significant consequences on health and

lifestyle, are typically experienced in these dynamic situations (Cavanagh et al.

1992; Horak et al. 1989b). To better understand why balance deficits occur in

these dynamic situations, knowledge about the fundamental strategies used by

the central nervous system (CNS) for coordinating the control of balance and

voluntary movements is needed. This information will provide a framework for

understanding balance deficits and the creation of novel rehabilitation

programs.

2

1.1 Scientific Background

Many of our daily activities involve reaching towards objects in our

extrapersonal space while maintaining upright stance. This apparently simple

task actually involves two divergent goals that must be coordinated by the

nervous system (Hess 1943; Massion et al. 2004). Specifically, the CNS must,

on the one hand, plan the trajectories of the limb segments to the goal

(movement), and on the other, maintain the stability of the limbs and balance of

the whole body (posture) (Horak and Macpherson 1996; Massion 1992;

Massion 1998). This duality between posture and movement, and their

underlying coordination, is non-trivial, and has been stressed since the early

studies of Babinski (1899), Hess (1943), Bernstein (1967), and several others

(Cordo and Nashner 1982; Massion 1992; Nardone and Schieppati 1988).

Moreover, the challenge of controlling these two divergent tasks is further

complicated by the need to control the many limbs, joints, and muscles

involved in a specific motor behaviour in the context of a given set of task

constraints. This problem is commonly referred to as the degree of freedom

(DoF) problem (Bernstein 1967). Consequently, it is thought that the CNS must

somehow simplify the complex task of controlling balance and movement with

suitable strategies so that skillful motor performance is achieved.

Due to the mechanical properties of the musculoskeletal system,

movement of any limb segment causes a shift in the position of the centre of

mass (CoM) that is disturbing to equilibrium (Hollerbach and Flash 1982;

Winter and Eng 1995). Furthermore, goal-directed, or focal, movements

involve complex spatial and temporal patterns of muscle activity in multiple

muscles throughout the body that result in net reaction forces that may also

have a perturbing effect on the body (Bouisset and Zattara 1981; 1987; Brown

and Frank 1987; Crenna et al. 1987; Friedli et al. 1988). To compensate for

these disturbances to equilibrium arising from the voluntary component of

actions, the CNS relies on postural muscle activity to stabilize the body (Horak

and Macpherson 1996; Massion 1992).

3

In the case of a voluntary movement, the CNS is able to predict the

mechanical consequences of the goal-directed movement and organizes the

appropriate muscle activity in the supporting limbs to compensate for the

disturbance (Bouisset and Zattara 1987; Cordo and Nashner 1982; Massion

1992). These postural adjustments are planned in advance of the movement and

are programmed using a feedforward mode of control (Gahery and Massion

1981; Massion 1992). In contrast, when balance is unexpectedly perturbed, the

CNS uses afferent information from the visual, vestibular and somatosensory

systems to shape feedback-based automatic postural responses (APRs) to

restore equilibrium (Nashner 1977).

Studies investigating postural responses to unexpected perturbations of

quiet stance have revealed much about the underlying control of posture. In

particular, specific strategies for controlling balance have been identified and

are thought to serve as a general mechanism by which the CNS simplifies the

coordination of motor behaviour (Ting 2007). For example, in response to

unexpected translations of the support surface in multiple directions, both cats

and humans restore equilibrium by using a ‘force-constraint’ strategy

(Macpherson 1988a), which involves exerting forces at the ground by each

supporting limb in only one of two directions irrespective of the direction of

translation (Henry et al. 1998a; Macpherson 1988a). In contrast, the

corresponding postural muscle activity is characterized by broad tuning across

perturbation direction, with maximal activity typically occurring for a single

characteristic direction (Carpenter et al. 1999; Henry et al. 1998b; Macpherson

1988b). More recent investigations of the coordination of muscle activity

underlying the feedback postural responses have revealed that the CNS appears

to simplify the task of balance control by constraining muscles to be activated

in fixed groups, or synergies (Ting and Macpherson 2005; Torres-Oviedo et al.

2006; Torres-Oviedo and Ting 2007). According to this model, the CNS sends

a command to activate a group of muscles, rather than controlling each muscle

individually. These muscle synergies correlate to the endpoint force vector

4

exerted under each supporting limb (Chvatal et al. 2011; McKay and Ting

2008; Ting and Macpherson 2005).

Muscle synergies are proposed to be a mechanism by which task-level

commands, such as CoM stabilization, are translated into execution-level

commands, which are the spatial and temporal patterns of muscle activity

(Chvatal et al. 2011; Ting 2007; Ting and Mckay 2007). Furthermore, muscle

synergies provide an attractive solution to the problem of controlling the

overabundant DoF associated with the multi-segmental arrangement of the

human body (Chiel et al. 2009; Ting 2007). It is not known, however, whether

similar strategies exist for feedforward postural control, in which postural

commands are organized in a predictive manner using knowledge of the body’s

dynamics and the external world (Bouisset and Zattara 1981; 1987).

During the execution of voluntary movements in the standing position,

preparatory anticipatory postural adjustments (pPA) occur prior to the onset of

the movement and serve to either stabilize the body or initiate movement,

whereas associated anticipatory postural adjustments (aPA) occurring during

the movement ensure a stable transition from one body configuration to

another. This postural activity is programmed in feedforward (Massion 1992)

and relies on prior knowledge of the arm and body dynamics (Bouisset and

Zattara 1981; 1987). Furthermore, anticipatory muscle activity in the supporting

limbs preceding a voluntary movement suggests that the CNS relies on a

predictive estimate of the future state of the body (Davidson and Wolpert

2005). This feedforward postural muscle activity is typically studied in

paradigms where the final goal of the movement is planned and therefore the

disturbance due to the movement can be predicted. For example, during a

planned voluntary movement, such as an arm raise, it is possible to predict the

dynamic consequences of the focal movement on equilibrium and therefore

plan the commands for posture accordingly (Massion 1992). However, if the

goal of the movement changes after the movement has been initiated, it is not

known how the CNS modifies postural control in relation to the online focal

correction in order to ensure that the target goal is met and equilibrium is

5

maintained. Investigating this question will provide fundamental insights about

how posture and movement commands interact, which has important

implications for understanding the deficits of balance control observed in

clinical settings.

1.2 Rational

To date, most studies investigating the control of posture during the

execution of voluntary movement have focused on either the role of the

anticipatory postural adjustments (APAs) preceding the movement or on the

temporal structure of these APAs with respect to the focal movement.

Consequently, there is still a lack of detailed information regarding the

organization of these postural adjustments when voluntary movements are

executed in multiple directions. Knowledge about how postural muscle activity

is coordinated with respect to movement direction is necessary to understand

how the nervous system may simplify the task of coordinating posture and

movement. Furthermore, no study to date has specifically addressed the online

control of posture in situations where the task goal changes after the movement

is initiated. These issues need to be clarified, both for a fundamental

understanding of the brain and motor system, as well as for a better

understanding of the postural deficits observed in several pathologies and

disease.

1.3 General Aim

As highlighted above, several specific questions remain about how

posture is controlled in relation to voluntary movement. Thus, the general aim

of the present thesis was to develop a greater understanding of the feedforward

adjustments of posture during voluntary movement. This was achieved with a

series of studies that (1) examined how postural adjustments are organized for

reach movements executed in multiple directions, and (2) investigated how

muscle activity for postural and movement related components of reaching

6

tasks are organized in relation to one another. This thesis was meant to go

beyond merely describing the role of feedforward postural control, but to

examine the spatial and temporal patterns of postural activity in order to make

supported inferences about how these postural adjustments may be organized

by the CNS.

1.4 Scientific Objectives and Hypotheses

Studies of postural control during voluntary movements in humans and

cats have provided evidence for the hypothesis that central commands for

maintaining balance interact with those for controlling movement (Massion

1992). Within this general framework, the main goal of this thesis is to explore

the organization of feedforward postural mechanisms in relation to voluntary

reaching movements performed in humans while standing. Specifically, this

thesis presents a series of experiments to (1) characterize the strategies adopted

for postural adjustments preceding and accompanying reaching movements in

multiple directions in standing (Specific Aim 1, SA1); (2) determine whether

muscle synergies explain the coordination of feedforward postural muscle

activity (SA2); and (3) explore the nature of postural control associated with

online corrections of reaching movements executed while standing (SA3). This

was achieved in Chapters 4, 5, and 6, respectively. Together, these studies

provide an important contribution to our knowledge about how the CNS

coordinates the control of posture and movement. To perform these

experiments, an experimental setup was designed to measure muscle activity,

kinetics and kinematics during whole-body reaching movements in multiple

directions, as detailed in Chapter 3. A description of the general objectives of

each study are presented below.

In Chapter 4, I quantify the strategies adopted by humans for

controlling posture during multidirectional reaching movements performed in

standing. I investigate whether predictive feedforward postural adjustments for

reaching are characterized by similar spatial organization of horizontal forces

exerted at the ground and muscle activity to those seen in reactive postural

7

control, despite their different modes of neural control. In Chapter 5, I extend

the findings of the previous chapter by identifying muscle synergies to explain

the spatial and temporal coordination of feedforward postural muscle activity.

Together, Chapters 4 and 5 provide the first characterization of feedforward

postural adjustments with the aim of drawing parallels with postural strategies

adopted for feedback postural control. Finally, in Chapter 6, I provide the first

examination of online postural control in relation to corrections in arm

movements resulting from visual perturbations of reaching while standing.

Specifically, I determine the mode of control for posture in relation to the signal

for the arm movement. Overall, the studies presented in this thesis further our

knowledge about feedforward postural control and contribute novel insight

about the coordination of posture and movement.

8

9

Chapter 2

Review of Literature

Imagine reaching forward to grasp an object located just beyond arm’s

length while standing. Most healthy people perform this skill with ease and

efficiency, many times a day, without any conscious thought to the underlying

control mechanisms. However, this seemingly simple behavior is complex, and

requires highly coordinated contributions between many structures distributed

throughout the CNS and musculoskeletal system that ensure the maintenance of

balance while performing the task. The underlying complexity in controlling

posture and movement arises from the multi-segmental structure of the body,

which affords the CNS significant redundancy and flexibility on the one hand,

but is also considered to be a computational challenge for the central controller.

Furthermore, the CNS must also consider how to accommodate the stability

requirements in the face of the disturbance of balance associated with the goal-

directed component of a movement. To simplify control, it is thought that the

CNS relies on neural strategies to transform task-level goals into appropriate

motor commands that specify appropriate muscle activation patterns (Ting and

Mckay 2007).

To establish a framework for understanding how the CNS coordinates

the control of posture and movement, the following review of literature will

first provide an overview of the underlying neurophysiology for the control of

movement and posture, with an emphasis on those circuits involved in the

coordination of movement and posture. Then, the biomechanical principles and

sensorimotor contributions governing the control of postural equilibrium and

orientation will be presented. Next, a discussion of the different modes of

postural control (feedback and feedforward) will follow. Finally, current

models of predictive motor control will be discussed.

10

2.1 How are voluntary movements and posture controlled?

The ability to move our body, limbs, head and eyes in order to

communicate, gesture and navigate our environment, while maintaining posture

and balance, is dependent upon the organization and complex interactions of

the motor and sensory systems (Ghez and Krakauer 2000). Successful planning

and execution of purposeful movement is largely facilitated by the hierarchical

organization of the sensorimotor systems distributed throughout the brain and

spinal cord. For example, the spinal cord contains the local circuitry that form

the building blocks used for both reflexive and goal-directed motor behaviours

(Rossignol et al. 2006). Higher levels of the CNS, such as areas of the

brainstem and motor cortex, provide descending modulation, which then

specify how the lower circuits are activated according to the task goals and

environmental context (Drew et al. 2004; McVea and Pearson 2009).

2.1.1 The neuroanatomical basis of movement execution

The production of movement involves four distinct, yet highly

interrelated neural subsystems. Each provides unique contributions to the

overall performance and control of movements. The four systems include the (i)

the motor neurons located in the spinal cord and brainstem, (ii) the motor

neurons of the brainstem and motor cortex, (iii) the cerebellum and (iv) the

basal ganglia (Purves et al. 2012).

The lowest level of the motor hierarchy involves the motor neurons

located in the spinal cord and brainstem. Their cell bodies are located in the

ventral horns of the grey matter of the spinal cord and the tegmentum of the

brainstem, and they synapse terminally on muscle fibers. These motor neurons

can influence the behaviour of their effector muscles directly via a pathway

involving its direct innervation of an alpha-motor neuron. Alternatively, these

motor neurons can serve as a relay for transmitting neural commands from

other motor neurons. As a result, the motor neurons of the spinal cord are

commonly referred to as the ‘final common pathway’ (Sherrington 1961). The

classification of motor neurons depends on the type of muscle fiber it

11

innervates. Alpha-motor neurons innervate extrafusal muscle fibers, which are

the skeletal fibers recruited for generating motion of the skeleton; and gamma-

motor neurons innervate intrafusal fibers, which are the contractile elements of

the muscle spindles (Pearson and Gordon 2000).

The second subsystem encompasses the motor neurons originating in

the motor areas of the frontal cortex and brainstem that subsequently synapse

on motor neurons and interneurons of the brainstem and spinal cord. These

motor neurons do not synapse directly with effector muscles, but rather shape

motor behaviour by modulating other motor neurons and the interneurons of the

spinal cord (Alstermark and Isa 2012; Ghez and Krakauer). These pathways are

important for the control of movement and posture. Notably, the corticospinal

tract is the primary descending pathway for planning and executing purposeful

movements (Ghez and Krakauer 2000). This tract may also influence posture

indirectly via collaterals that branch at different levels of the CNS (Kably and

Drew 1998; Massion 1992). Additionally, the reticulospinal tract has been

shown to have important contributions for the maintenance of balance (Drew et

al. 2002; Luccarini et al. 1990; Mori).

The cerebellum and the basal ganglia form the third and fourth

subsystems, respectively. Although these structures do not innervate muscle

fibers directly, they contribute to motor control via their modulation on the

activity of motor neurons. In particular, the cerebellum is thought to detect and

signal motor error (Desmurget and Grafton 2000), which is the difference

between the intended and actual movements. As such, the cerebellum is critical

for processes related to motor adaptation and long term learning (Bastian 2006).

The basal ganglia, which forms the fourth subsystem, is a collection of

structures located in the deep forebrain. The structures of the basal ganglia are

responsible for inhibiting unwanted movements and preparing the upper motor

neurons for the initiation of motor action (Mink 1996).

Together, these subsystems, along with the associated sensory systems,

form a complex hierarchical system that enables the performance of a vast

range of movements. For the purpose of the present discussion, only the

12

circuitry of the spinal cord, brainstem and motor cortex will be reviewed in

more detail.

2.1.2 Circuitry of the spinal cord provides a basis for coordinating

movement

Reductionist approaches to the study of motor behaviour provided

considerable insight into the circuitry of the spinal cord and how it contributes

to motor behaviour (Eccles and Sherrington 1930; Sherrington 1910). Using

reduced animal preparations involving the transection of the spinal cord or

brain at different levels (Sherrington 1909), or by deafferentation of a muscle of

interest, these studies identified many of the important components of the spinal

cord. Moreover, these studies have explained how the underlying circuitry of

the spinal cord forms the basis of coordinated motor behaviours, including

reflexes (Eccles and Sherrington 1930) and rhythmic behaviour such as

swimming (Grillner 2003) and locomotion (Rossignol et al. 2006).

A classically cited example to explain the spinal circuitry is the stretch

reflex (Pearson and Gordon 2000). The stretch (myotatic) reflex involves the

most basic elements of the spinal cord for producing a mechanical behaviour

without input from higher centers (Liddell and Sherrington 1924). For example,

when a muscle is passively stretched, this deformation is sensed by a muscle

spindle, which is a sensory receptor located within the muscle belly. Increases

in stretch in the muscle result in increased firing of the Ia-afferent sensory fiber,

which in turn excites the alpha-motor neuron of the muscle being stretched via

a direct synapse as well as the alpha-motor neurons of the synergistic muscles.

The Ia-afferent also inhibits the activity of the alpha-motor neuron innervating

the antagonist muscle. The net result of this activity is to resist the stretch on

the muscle by contracting the stretched muscle, thus generating force in that

muscle to cause flexion of the joint crossed by the involved musculature

(Pearson and Gordon 2000). This mechanism illustrates how complex motor

behaviour involving more than one muscle can be achieved by a simple

negative feedback loop mechanism.

13

The neural networks housed in the spinal cord are known as central

pattern generators (CPGs). They are capable of controlling the timing and

coordination of muscle activation patterns appropriate for the environment,

even if those environmental conditions change (Rossignol et al. 2006). CPGs

form the building blocks responsible for generating locomotor or rhythmic

behaviour in many species, including the lamprey (Grillner 2006), cats

(Rossignol et al. 2006), and frogs (Tresch et al. 1999). For example, in the

lamprey, swimming behaviour is achieved by coordinated activation and

inhibition of muscle segments along the length of the animal’s body to produce

sinusoidal motion (Grillner 2006). The neural mechanisms responsible for this

coordinated activity has been studied extensively (Grillner 2006).

The modular organization of the spinal cord is not limited to lower

vertebrates, but rather appears to be fundamental for controlling motion of the

skeleton in a variety of behaviours in several species (Tresch et al. 1999). For

example, several different types of behaviour can be elicited in a spinalized frog

by recruiting the same neural circuits (Cheung et al. 2005; Tresch et al. 1999).

Also, in a spinalized cat, locomotor behaviour can be elicited with the

appropriate sensory input and sufficient postural support to compensate for the

lack of postural tone (Rossignol et al. 2002). This suggests that higher neural

centres provide important modulatory control on the activity of the lower spinal

circuits. However, spinalized cats lack the capacity to maintain balance in

response to unexpected perturbations (Macpherson and Fung 1999). Whether

this is due to the trauma endured by the spinal cord or due to the lack of input

from higher centers remains to be determined (Honeycutt and Nichols 2010). In

support of this view, Schepens and Drew (2006) have proposed that the same

networks identified for the control of locomotion are also recruited in the

control posture.

2.1.3 Somatotopic organization of spinal cord

The spatial distribution of the motor neuron pools in the spinal cord has

been mapped by injecting different muscles with retrograde tracers that label

14

the cell bodies of the motor neurons that innervate that muscle (Levine et al.

2012). Subsequently, histological analyses were used to determine the specific

spatial mappings that exist along both the superior-inferior and medial-lateral

aspects of the spinal cord and brain stem.

Motor neuron pools innervating the upper limbs are found in the

cervical enlargement of the spinal cord, whereas the lower limbs are innervated

by motor neurons in the lumbar regions of the spinal cord (Ghez and Krakauer

2000). Furthermore, along the medial-lateral aspect, motor neuron pools are

organized such that the axial muscles, which are important for postural control,

are located in the medial and anterior regions of the ventral horn of a spinal

cord segment. Accordingly, motor neurons located more laterally within the

ventral horn innervate muscles placed more laterally in the body. Finally, the

most distal muscles of the body extremities, such as the digits, are innervated

by motor neurons that are the most laterally placed from the midline of the

spinal cord. Thus, an apparent functional grouping of musculature from axial,

proximal and distal can be observed as a result of the somatotopic map present

even at the lowest level of organization within the CNS (Levine et al. 2012).

The somatotopic organization of the local circuitry of the spinal cord

along the longitudinal axis also reflects functional differences of the networks

involved in controlling posture and those for skilled movement. In particular,

the local circuits in the medial regions of the intermediate zone synapse on the

motor neurons in the medial ventral horn. The axons involved in these circuits

have some projections that span multiple spinal segments enabling the

coordination of upper and lower limbs, while others synapse the length of the

spinal cord to assist in controlling posture. Some still cross the midline in the

commissure of the spinal cord providing a means for coordinating left and right

axial muscles (Ghez and Krakauer 2000). The local circuitry for the lateral

motor neurons, however, is increasingly differentiated with projections

contained mainly to the ipsilateral side of the spinal cord and spanning no more

than five spinal segments (Ghez and Krakauer 2000). This connectivity is

important for the highly differentiated control of distal muscles recruited for

15

skilled motor behaviour, such as independent control of the digits during

manipulation tasks (Alstermark and Isa 2012). The differences in the local

circuitry of the medial and lateral networks, as well as in the somatotopic

organization of the motor neurons in the ventral horn results in functional

differences in the control of muscles for either postural or goal-directed tasks.

The anatomical organization of the spinal circuits and their effector muscles

provide an anatomical substrate for understanding how the CNS might achieve

the complex coordination of many muscle groups across various parts of the

body.

2.1.4 Anatomical organization of the descending pathways for the control

of movement

The second subsystem consists of the upper motor neurons located in

the brainstem and cerebral cortex. These supraspinal networks provide

important modulation of the lower circuits in the brainstem and spinal cord

upon which they synapse via direct and indirect connections (McVea and

Pearson 2009). These descending pathways are divided into the lateral and

medial systems, according to differences in their spatial organization in the

spinal cord and functional connectivity (Drew et al. 2002). The lateral pathway,

which encompasses the lateral corticospinal tract and the rubrospinal tract, is

located laterally within the spinal cord and is primarily involved in movement

initiation and specifying the complex temporal coordination of muscle activity

for skilled voluntary movements (Ghez and Krakauer 2000). In contrast, the

medial pathways, which include the reticulospinal and vestibulospinal tracts,

have fairly vast innervation patterns of the axial and proximal musculature and

are critical for regulating posture (Drew et al. 2002) and orienting the head,

trunk with respect to vestibular, somatic, auditory and sensory information

(Ghez and Krakauer 2000).

16

2.1.4.1 Lateral descending pathway

The primary tract of the descending pathway is the corticospinal tract.

It, along with the corticobulbar tract, originate from the pyramidal cells of layer

5 of the motor cortex (Areas 4 and 6 of the frontal lobe) and terminate in the

spinal cord and brainstem nuclei, respectively. The corticospinal tract synapses

with motor neurons and interneurons of the brainstem and spinal cord, whereas

the corticobulbar tract projects to brainstem nuclei responsible for the control of

the cranial muscles.

As the corticospinal and corticobulbar tract descend the nervous system,

both have collaterals that branch at various levels to innervate the cranial

nuclei, the reticular formation, the red nucleus, and the pons. At the caudal level

of the medulla, only corticospinal tract (CST) axons remain. At this point,

approximately 90% of the CST axons cross the pyramidal decussation to form

the lateral CST on the contralateral side of the spinal cord and modulate the

lateral spinal motor neurons. The remaining 10% of CST axons continue to run

ipsilaterally or bilaterally, and form the ventral (anterior) CST, which is a part

of the ventral descending system. This portion of the CST originates in the

dorsal and medial regions of the motor cortex and terminates mainly on the

circuits serving axial and proximal musculature. Interestingly, this group of

CST neurons also give rise to the projections terminating in the reticular

formation, providing them with a privileged role in the control of posture.

Overall, the terminal distribution of the CST axons suggests functionally

different roles of the lateral and ventral divisions of the CST, whereby the

lateral CST provides important contributions to the fine control of the upper

extremities and the ventral CST for the control of posture (Ghez and Krakauer

2000; Purves et al. 2012).

The rubrospinal tract arises from the red nucleus in the midbrain and

immediately crosses to the contralateral side of the brain to descend via the

brainstem and innervate spinal motor neurons at several levels of the spinal

cord (Ghez and Krakauer 2000). Similar in function to the CST, the rubrospinal

tract is an indirect pathway for the control of voluntary movements. In

17

primates, however, much of the function of the rubrospinal tract has been

assumed by the CST (Ghez and Krakauer 2000).

2.1.4.2 Medial descending pathway

The diffuse action of the medial descending system is achieved by the

action of four descending pathways: the vestibulospinal tract, the tectospinal

tract and pontine (medial) reticulospinal tract and medullary (lateral)

reticulospinal tract. These pathways arise from various nuclei distributed

throughout the brainstem, including the vestibular nuclei, the tectum, the pons

and medulla, respectively, and project to the ventromedial regions of the spinal

cord gray matter. Together, they are critical for maintaining balance and

orienting the body and gaze (Drew et al. 2002).

For example, the vestibular nuclei give rise to the vestibulospinal

pathway, which has direct connections to the spinal cord, and is therefore able

to rapidly recruit the appropriate postural networks in response to a disturbance

of balance detected by the vestibular apparatus (McVea and Pearson 2009). The

tectospinal tract originates from the superior colliculus, which integrates visual

input received directly from the retina with somatosensory and auditory

information, and terminates in the upper segments of the spinal cord (Purves et

al. 2012). Consequently, it has been proposed that the tectospinal pathway may

have a role in the orientation of gaze. Finally, the reticular formation of the

brainstem forms a complex network of circuits originating in either the pons or

the medulla to form the pontine (medial) reticulospinal tract and the medullary

(lateral) reticulospinal tract. These two pathways have been shown to be

involved in a vast number of functions, including breathing, sleeping, posture

and locomotion (McVea and Pearson 2009).

18

2.1.4.3 Anatomical organization of reticular formation of the brainstem

relevant for postural control

Amidst the many functions of the reticular formation of the brainstem, it

has notable contributions in the control of posture (Drew et al. 2002), likely as a

result of its diffuse projections throughout the CNS. Although both the medial

and lateral reticulospinal tracts have been shown to be involved in organizing

the spatial and temporal coordination of the activation patterns of the trunk and

limb muscles (Drew et al. 2002), the medial pathway appears to have a more

significant role (Brustein and Rossignol 1998). In part, this can be explained by

differences in the connectivity of the two pathways. The medial brainstem

pathways terminate in the ventromedial areas of the spinal cord, providing

important modulatory effects on the axial and proximal musculature. These

pathways are vast in their projections, enabling the modulation of several

functionally related groups of motor nuclei (Drew et al. 2002). In contrast, the

lateral brainstem pathways project to the lateral areas of the spinal cord and are

consequently involved in the control of the distal muscles of the extremities

required for skills such as reaching and grasping (Ghez and Krakauer 2000).

In summary, knowledge of the basic neuroanatomical pathways related

to the initiation and execution of movement and the control of posture is crucial

for understanding how the CNS coordinates the postural and focal demands of a

task. Notably, the primary descending pathway, the corticospinal tract, has

collaterals that branch to the vestibular and reticular nuclei of the brainstem

prior to the decussation (Kably and Drew 1998). These areas are known to have

privileged involvement in controlling the body’s posture and equilibrium (Drew

et al. 2004). Therefore, it is plausible that these circuits provide the basis for the

neural computing required for integrating these two dichotomous aspects of

motor behaviour.

19

2.1.5 Integration of central commands for the global planning of movement

and posture

Cortically-initiated voluntary movements are preceded and accompanied

by postural adjustments that ensure balance is maintained (Massion 1992).

Questions related to the mechanisms through which these two aspects of motor

behaviour are integrated, and the localization of the networks responsible for

integrating posture and movement have received considerable attention in the

literature (Massion 1992). In the following paragraphs, the modes of control for

coordinating posture and movement will be explored, followed by a discussion

of the localization of the neural substrates involved in this coordination.

2.1.5.1 Modes of control for the integration of posture and movement

In consideration of several studies in humans and animals, Massion

(1992) proposed a classical theoretical model to explain how posture and

movement commands may be organized in the CNS. In this framework, two

possible modes of control have been identified (see Fig 16, (Schepens and

Drew 2004).

In the first, posture and movement are controlled via parallel pathways.

Here, it is postulated that posture and movement are controlled independently;

with separate descending neural commands that target either the postural

network or the movement related circuits. The feedforward postural

adjustments preceding movements are typically attributed to this mode of

control, such as during arm (Bouisset and Zattara 1981; 1987) or leg

movements (Mouchnino et al. 1992) in humans and reaching movements in cats

(Schepens and Drew 2003).

In general, temporal decoupling of the signal for the postural adjustment

and the movement is cited as evidence that these two commands are controlled

independently (Massion 1992). For example, the latency of the postural

adjustment preceding an arm movement is modulated as a function of the time

constraints imposed by the task. When the movement is self-paced, postural

activity occurs prior to the movement whereas in a reaction time paradigm,

20

postural and focal activities occur simultaneously (Lee et al. 1987).

Furthermore, the latency of the feedforward postural adjustment prior to the

movement onset increases as the load to be lifted increases (Bouisset and

Zattara 1981). It is hypothesized that the movement command is inhibited until

the postural system reaches a desired reference point optimal for the movement

execution (Cordo and Nashner 1982; Massion 1992). However, the mechanisms

of this inhibition and the nature of the command signals remain largely

unknown.

Alternatively, posture and movement commands may be controlled

hierarchically by a shared descending command that modulates the postural

networks via collateral branches (Gahery and Massion 1981; Gahery and

Nieoullon 1978). This mode of control appears to be used for organizing the

feedforward postural adjustments accompanying the movement (Massion

1992), such as during a bimanual load-lifting task (Paulignan et al. 1989). For

example, when the onset of the postural adjustment is time-locked to the focal

movement, it suggests that posture and movement commands are controlled

together within a hierarchical organization (Gahery and Nieoullon 1978;

Paulignan et al. 1989).

The theoretical model proposed by Massion (1992) has been

substantiated and extended by a series of behavioural and electrophysiological

studies examining the nature of the control signals for posture and movement

during reaching movements in the cat (Schepens and Drew 2006; 2004; 2003;

Yakovenko and Drew 2009). These findings will be discussed in detail in the

next section, but overall, these studies provide evidence that both hierarchical

and parallel modes of control are used by the CNS. Which mode of control is

adopted, however, may depend mainly on the role of the postural adjustment

and the nature of the voluntary task (Massion 1992).

2.1.5.2 Central organization of feedforward postural adjustments

Beyond the neuroanatomical evidence that points to the involvement of

several cortical and subcortical structures in coordinating posture and

21

movement, evidence from lesion, microstimulation and single-unit recording

studies have elucidated many of the central pathways responsible for the

coordination of posture and movement. There is general agreement that the

supraspinal regions of the CNS are involved in planning and initiating

feedforward postural adjustments (Gahery and Nieoullon 1978; Horak and

Macpherson 1996; Mackinnon et al. 2007; Massion 1992; Viallet et al. 1992;

Yakovenko and Drew 2009), whereas the brainstem is thought to be vital for

ensuring the appropriate scaling of the postural adjustments in relation to the

movement by integrating information from several cortical and subcortical

structures (Drew et al. 2004). Finally, it has been hypothesized that the

descending commands recruit the same CPG networks used in locomotion for

coordinating the muscle activity of the postural response (Honeycutt et al.

2009; Schepens and Drew 2006).

Evidence that these postural adjustments have contributions from

descending pathways arising from the cortex was demonstrated by Gahery and

Nieoullon (1978), who performed microstimulation of the motor cortex in the

cat to induce flexion movements of a limb. They found that the induced flexion

movement, whether it was of the fore- or hindlimb, was associated with a

decrease in the loading force under the limb diagonal to the flexed limb and a

concurrent increase in the loading forces of the other two limbs (termed

‘diagonal strategy’). Analysis of the latencies of the force changes revealed that

the postural adjustments for hindlimb flexion clearly preceded the limb flexion,

whereas those for forelimb flexion occurred concurrently or just after the

loading change in the moving limb. These data were interpreted to indicate that

the postural adjustments accompanying the flexion movements are centrally

driven and not the result of feedback driven EMG from peripheral afferents,

although feedback from the periphery may be used at a later stage of the

postural adjustment (Gahery and Nieoullon 1978). Moreover, it was

hypothesized that the pathways for the movement command send collaterals to

the brain stem or spinal cord to modulate the lower level networks responsible

for postural control (Gahery and Massion 1981; Massion 1992).

22

A cortical contribution to the control of posture during intentional

movement is also supported by studies in humans (Mackinnon et al. 2007;

Viallet et al. 1992). Using a bimanual load-lifting task (Dufosse et al. 1985;

Hugon et al. 1982; Paulignan et al. 1989) in patients with lesions in the medial

frontal areas, Viallet and colleagues (1992) demonstrated that both hemispheres

of the cortex are required for producing appropriate postural adjustments during

voluntary movements to maintain the stability of the limb. In this task, a load is

placed on the subject’s forearm and they are required to remove it by

voluntarily lifting the load with the other arm. In control subjects, the position

of the loaded arm remains fairly stable, even as the load is removed. Analysis

of the muscle activity revealed anticipatory inhibition of the loaded forearm’s

flexors that was time locked to the voluntary contraction of the muscles of the

moving segment (Paulignan et al. 1989; Viallet et al. 1992). However,

hemiparetic patients lacked appropriate anticipatory postural adjustments

(APAs) in the arm contralateral to the lesion (Viallet et al. 1992). The authors

suggested that the supplementary motor areas and associated motor cortical

areas contralateral to the loaded arm are necessary for organizing the phasic

anticipatory postural adjustment, whereas the cortical motor areas contralateral

to the moving limb recruit the networks for the focal muscle activity and, via

collaterals, the networks for the postural adjustments accompanying the

movement.

Also, Mackinnon and colleagues (2007) demonstrated that the

amplitude of both the postural activity preceding step initiation and the focal

muscle activity for stepping were increased in a time-locked manner as a result

of TMS applied to the leg area of the motor cortex in humans. These results

further suggest that the commands for posture and movement may be shared.

Finally, evidence from neural recordings in the motor cortex during

reaching movements in the cat have shown a specific contribution of the

descending tracts from the motor cortex to the postural adjustments preceding

movement onset (Yakovenko and Drew 2009). Here, cats performed reaching

movements in an instructed delay task while standing. The discharge of several

23

pyramidal tract neurons was shown to be time-locked to the cue for the ‘GO’

signal and not to the onset of the goal-directed component of the reaching

movement. Furthermore, the activation latency of these neurons was correlated

to the onset of the preparatory postural activity, indicated by a change in

vertical pressure under the supporting limbs. Finally, the discharge

characteristics of the recorded cortical neurons were identical to those

previously recorded in the pontomedullar reticular formation (PMRF)

(Schepens and Drew 2006; 2004; Schepens et al. 2008), a structure known to

have important functions in postural control (Drew et al. 2004) rather than in

movement planning. These findings are consistent with previous lesion and

stimulation studies (Perfiliev 2005) that have suggested that the motor cortex

contributes to the descending signal for anticipatory postural adjustments, but

suggests that it’s contributions are not involved in the global planning of

posture and movement processes (Yakovenko and Drew 2009).

Beyond the cortex, the brainstem reticular formation has been attributed

a critical role in the integration of neural commands for posture and movement

(Schepens and Drew 2003; Schepens et al. 2008). Lesion studies in cats and

primates have shown that damage to the reticulospinal and vestibulospinal

tracts result in deficits in balance control whereas fine control of the distal arm

is maintained (Lawrence and Kuypers 1968). Also, inactivation of the pontine

reticular formation by a cholinergic agonist during cortical stimulations to elicit

feedforward postural adjustments result in deficits in the postural response

(Luccarini et al. 1990). Finally, selective lesions of the medial reticular tract in

cats results in deficits in posture and weight support during locomotion, where

the magnitude of the postural deficit is related to the severity of the lesion

(Brustein and Rossignol 1998).

The importance of the brainstem to the control of balance is supported

by studies of reactive postural control. For example, decerebrate cats are

capable of restoring balance following unexpected perturbations of balance

with similar strategies to those observed in intact animals (Honeycutt et al.

2009; Honeycutt et al. 2012; Honeycutt and Nichols 2010), suggesting that the

24

primary networks for postural control are located at lower levels of the motor

hierarchy. However, for postural responses to be appropriately scaled in time

and magnitude to the parameters of the perturbation, descending input from the

cortex is required (Drew et al. 2004). For example, cats with lesions to the

spinal cord are able to restore balance following a perturbation, but lack the

precise spatial organization of muscle activity in several muscles that is

typically observed in controls (Macpherson and Fung 1999), lending support to

the view that the brainstem has a privileged role in equilibrium. However, some

authors have argued that the deficits in postural control observed in spinalized

animals may be due to the trauma of the lesion and subsequent reorganization

of signaling pathways rather than due to the loss of descending input from the

brainstem or cortical areas (Honeycutt and Nichols 2010).

Recently, a series of electrophysiological studies recorded neurons in

the brainstem reticular formation during reaching movements and unexpected

perturbations in cats (Schepens and Drew 2006; 2004; 2003; Schepens et al.

2008; Stapley and Drew 2009). These have provided considerable insight about

how and where commands for posture and movement are controlled within the

CNS. Specifically, three populations of cells in the PMRF were identified that

were related to either the postural adjustment preceding the movement,

accompanying the movement, or the movement itself (Schepens and Drew

2004; Schepens et al. 2008). Furthermore, discharge of PMRF neurons occurred

for reaches performed with either limb, although was increased for reaches

made with the limb ipsilateral to the recording site (Schepens and Drew 2006).

Quantitative analyses in the form of regressions and spike triggered averaging

(STA) revealed that the bilateral signal is asymmetric in nature and the final

descending signal is modified, or gated, depending on which limb performs the

reaching movement (Schepens and Drew 2006). Interestingly, changes in the

discharge of the neurons of the PMRF were also observed when the support

surface under one of the limbs was unexpectedly removed (Stapley and Drew

2009). Together, these studies demonstrate that the PMRF contributes

significantly to the coordination of posture during both goal-directed

25

movements and unexpected perturbations of the limb. These findings suggest

that the PMRF gates neural signals for both feedback and feedforward signals

and is the site of integration of postural control organized via these two

mechanisms.

2.2 Postural Control

2.2.1 Biomechanical requirements for equilibrium control

Humans are faced with a unique set of equilibrium constraints as a

result of their upright, bipedal stance. In particular, the CNS must deal with a

relatively high position of the center of mass (CoM) within a narrow base of

support (BoS), which consequently affords a limited region of stability (Hayes

1982). To maintain stability and equilibrium during quiet standing and while

executing complex multi-joint movements, the human nervous system must

balance all external and internal forces acting on the body (Horak and

Macpherson 1996; Massion 1992). The forces acting on the body include

external constraints associated with the environment and task, such as the force

of gravity, the reaction forces from the supporting surfaces, and any imposed

accelerations or obstacles in the environment (Massion 1992). In addition,

internal constraints refer to those constraints that arise within the body, such as

the geometry and inertial characteristics of the body segments, and the internal

forces associated with muscle contractions (Massion 1992). For the skillful

execution of movement, the CNS must choose an optimal neural strategy in

consideration of the task and biomechanical, and musculoskeletal constraints

that optimally balances all these forces in order to meet the task goals (Horak

and Macpherson 1996; Massion 1992).

Equilibrium is achieved by maintaining the position of the body’s

CoM, defined as the weighted sum of the body’s segments, within a relatively

small region of stability afforded by the supporting limbs, known as the BoS

(Winter 2009). The tight regulation of the CoM for balance control is illustrated

by the phenomenon of body sway that occurs naturally in quiet stance. In quiet

26

upright stance, the control of balance can be modeled approximately as an

inverted pendulum (Pai and Patton 1997; Winter et al. 1998). The net

acceleration on the body is determined by the moment of force resulting from

the perpendicular difference between line of gravity through the CoM and the

ground reaction force (GRF). The position of the centre of pressure (CoP;

defined as the point of application of the GRF (Horak and Macpherson 1996;

Winter 2009)) is actively modulated by the ankle musculature (Morasso and

Schieppati 1999), which in turn, results in adjustments of the distance between

the line of gravity through the CoM and the GRF (van Ingen Schenau et al.

1992). The perpendicular difference between these two resultant force vectors

creates a moment of force that causes a net acceleration of the whole body

about the ankle joint (Winter 2009). In summary, the CNS actively modulates

the point of application of the external GRF, which corresponds to the CoP, to

achieve the appropriate resultant moment of force that will cause the desired

whole body acceleration. In addition to sway, this control mechanism has been

demonstrated for a number of dynamic tasks, including cycling (van Ingen

Schenau et al. 1992), whole body lifting (Commissaris et al. 2001; Toussaint et

al. 1995), and locomotion (Lepers and Brenière 1995).

2.2.2 Behavioural goals of the postural system

The act of maintaining balance both during quiet stance, and when

transitioning from one body configuration to another during voluntary

movements, engages several systems distributed throughout the CNS. Together,

these processes serve the goals of the postural system. In the context of goal-

directed motor behaviour, the postural system provides an essential foundation

for motor coordination by controlling the position of the CoM, thus providing

stability for transitioning between body configurations and orienting the body

segments with respect to each other, the environment or both (Horak and

Macpherson 1996).

Broadly, the behavioural goals of the postural system can be categorized

as being involved either in specifying the orientation or in maintaining

27

equilibrium of the body and its segments. Specifically, postural orientation

refers to those processes that maintain the alignment of the body segments with

respect to themselves and to the environment (Horak and Macpherson 1996;

Massion and Woollacott 2004). For any given postural configuration, postural

variables such as the alignment of the head, trunk, and the geometry of the

limbs are stabilized according to the task and context (Horak and Macpherson

1996). In contrast, postural equilibrium ensures that all forces acting on the

body are balanced so that the body remains balanced or moves in a controlled

fashion (Horak and Macpherson 1996). This is achieved with coordinated

postural strategies that respond to disturbances of balance resulting from either

internal or external perturbations.

2.2.3 Sensorimotor control of posture

Feedback from a number of sensory modalities is utilized by the CNS to

create a representation of the body in relation to its external environment

(Horak and Macpherson 1996; Mergner and Rosemeier 1998). In particular, the

visual, vestibular and somatosensory, which includes muscle proprioception,

joint and cutaneous afferents, systems each detects an ‘error’ signal relating the

amount of deviation of the body orientation from a reference point (Peterka

2002). These signals are then thought to be integrated and weighted according

to the task, context and availability of sensory information (Horak et al. 1990).

The integrated sensory input is subsequently used to calculate the orientation of

the head, trunk and limbs in space, providing an internal representation of the

body and its surroundings (Horak and Macpherson 1996; Mergner and

Rosemeier 1998).

The relative contributions of the different sensory modalities to the

overall control of postural orientation and equilibrium have been investigated

experimentally with a number of paradigms. Typically, these paradigms involve

systematically manipulating each of the modalities and measuring the effects on

balance control in quiet stance (Diener et al. 1984b; Nashner et al. 1982), in

response to unexpected perturbations of balance (Horak et al. 1990) or during

28

voluntary movements (Macpherson et al. 2007). Overall, these studies have

shown that each sensory modality is unique and responds optimally for a given

behaviour (Horak, 1996 (Horak and Macpherson 1996; Horak et al. 1990;

Mergner and Rosemeier 1998).

For example, when pressure cuffs are applied to the ankles to limit the

input from somatosensory afferents arising from the feet, little effect on

postural sway is observed when vision is available. However, when subjects are

asked to close their eyes, increased sway at the hips is apparent (Diener et al.

1984b). Similarly, patients with vestibular loss are able to compensate for their

sensory loss when visual input is available. However, when vision and

somatosensory inputs are inaccurate, these patients experience a loss of

equilibrium (Nashner et al. 1982). In general, these studies suggest that

somatosensory and vestibular information have critical roles in controlling

balance during upright stance (Horak et al. 1990).

While the study of sway during quiet stance has provided important

insight into the sensorimotor processes of postural control, this approach is

limited in its ability to specify the mechanisms for dynamic postural control,

such as in response to an unexpected loss of balance or during the execution of

goal-directed movements while standing. Accordingly, investigators have

examined how sensory processes shape postural behaviour by quantifying the

postural responses to dynamic perturbations of balance when one or more

sensory inputs is disrupted in cats (Inglis and Macpherson 1995; Macpherson et

al. 2007; Stapley et al. 2002; Stapley et al. 2006) and humans (Horak et al.

1990). Overall, these studies have shown that not only is the detection of the

postural disturbance affected by sensory loss, but so is the selection and

organization of an appropriate postural strategy (Horak et al. 1990).

In particular, studies in cats have provided important insight about the

relative contributions of vestibular and somatosensory inputs. In one such

study, Inglis and Macpherson (1995) subjected cats to support surface

translations in the horizontal plane before and after bilateral labyrinthectomy.

Following the lesion, the postural strategies were similar in spatial and temporal

29

organization to the control responses even when executed in complete darkness.

However, increases in the amplitude of the responses were observed. These

results suggest that vestibular information is not critical for the organization of

the postural response, but rather influences its scaling. Somatosensory

information, however, appears to have a more important role in shaping the

temporal and spatial characteristics of the postural response (Maurer et al. 2001;

Stapley et al. 2002; Ting and Macpherson 2004).

The role of somatosensory information in shaping the postural responses

to unexpected perturbations of balance was examined specifically in cats given

large doses of pyridoxine to induce selective large-fiber deafferentation

(Stapley et al. 2002). The results of this study provide evidence that the large

afferent fibers carrying somatosensory information from the periphery are

responsible for the correct timing of the coordinated postural strategies that

restore equilibrium following an unexpected perturbation. Furthermore, Ting

and Macpherson (2004) demonstrated that somatosensory information is critical

in shaping the postural response by comparing postural responses to multi-

directional pitch/roll rotations and linear translations in the horizontal plane of

the support surface in freely standing cats. They found that the only reliable

parameter of the perturbation in predicting the organization of the postural

response was the ratio of the vertical loading force and horizontal slip force

components of the GRF. This signal presumably reflects the horizontal motion

of the CoM. Based on their findings, the authors proposed that cutaneous

sensors in the feet detect the change in force angle at the ground and

subsequently provide critical information regarding the nature of the

perturbation necessary for organizing the postural response (Ting and

Macpherson 2004). The importance of somatosensory cues from the feet for

controlling balance has also been demonstrated for humans (Maurer et al.

2001).

In summary, the postural control system encompasses all neural,

muscular and sensory processes related to maintaining the equilibrium and

orientation of the body (Horak and Macpherson 1996). The sensory modalities

30

are critical for providing cues about the orientation of the limbs and body in

relation to a reference point. The inputs are thought to be summed (Peterka

2002) and weighted according to the task and context (Jeka et al. 2006;

Mergner and Rosemeier 1998). In the event of the loss or disruption of a

sensory modality, redundancy across the sensory systems, as well as the ability

to re-weight the incoming sensory input, ensure that at least partial recovery of

a representation of the body schema is possible (Black et al. 1989).

2.2.4 The problem of motor redundancy

2.2.4.1 Degree of Freedom Problem

To control balance and posture while standing, the CNS must coordinate

the activity of many muscles crossing a large number of joints capable of

moving in multiple planes (Bunderson et al. 2010). As a result, there is

significant redundancy in the motor system for defining a unique pattern of

joint motion (Kuo 2005; Todorov 2004; Yang and Pai 2007) and muscle

activation patterns (Gottlieb 1998; Lockhart and Ting 2007) for a given motor

task. This classical problem of motor control was first formalized by Bernstein

(1967) and labeled the degree of freedom (DoF) problem (Bernstein 1967).

Faced with this redundancy, it is thought that the CNS must compute an

appropriate set of motor commands to generate muscle activity that is

appropriate for the task and environmental constraints, all while considering the

complex kinematic chains of the musculoskeletal system and the inertial

properties of the body’s segments (Chiel et al. 2009). How the CNS maps task-

level goals into coordinated patterns of muscle activity remains a topic of

considerable debate amongst researchers (Latash 2012).

A widely accepted hypothesis is that the CNS is organized to control

several functionally related DoF together, as modules (Bernstein 1967;

Krishnamoorthy et al. 2003; Krishnamoorthy et al. 2004; Ting 2007; Ting and

Mckay 2007). Within this framework, higher centers of the CNS serve to

constrain global parameters for movement, such as task-level goals, whereas

31

lower levels of the hierarchy specify the coordinated spatial and temporal

patterns of the muscle activations (Ting 2007; Ting and Mckay 2007). This

hierarchical organization is thought to decrease the complexity of control by

diminishing the dimensionality of the control problem (McKay et al. 2007;

McKay and Ting 2008), while providing an attractive mechanism to explain

how the CNS maps task-level goals into appropriate muscle activation patterns

(Chvatal et al. 2011; Ting 2007).

Support for a hierarchical organization of the motor system is evident

from several psychophysical, biomechanical and neurophysiological studies

that have demonstrated that task-level variables are often constrained with

greater rigidity than lower level goals (Chiel et al. 2009; Ting 2007). For

example, in a reaching task, end-point trajectory of the finger is more tightly

regulated than the angular displacements of the individual joint angles.

Furthermore, task level parameters, such as direction, force and velocity of the

hand during reaching, are encoded in the motor cortex in primates

(Georgopoulos 1986). Similarly, in human locomotion, changes in joint angles

are correlated, suggesting that the CNS constrains several joints together to

control CoM motion (Bernstein 1967; Cavagna et al. 1977). Together, these

findings provide support for the hypothesis that the CNS simplifies the control

of redundant systems by grouping DoFs together as ‘synergies’ or ‘modules’

and serve as a mechanisms to control task level goals (Chvatal et al. 2011;

McKay and Ting 2008; Ting 2007; Ting and Mckay 2007; Yakovenko et al.

2011). However, it remains to be determined at what level of the motor system

this redundancy is resolved.

2.2.4.2 Muscle synergies simplify the coordination of muscle activity

Recently, this model of hierarchical control has been developed further

in the context of muscle coordination. It is proposed that the CNS solves the

problem of redundancy at the level of the muscles through the flexible

recruitment of muscle synergies (Lockhart and Ting 2007; Ting 2007; Ting and

Mckay 2007). A muscle synergy is defined as invariant patterns of muscle

32

activation with fixed spatial scaling (Ting 2007; Ting and Macpherson 2005). It

is postulated that the nervous system specifies a neural command that

determines the temporal recruitment of a muscle synergy.

Within this framework, the overall muscle activity is expected to reflect

the linear sum of each of these synergies multiplied by their respective

activation coefficient, which are the purported neural commands (Ting 2007).

Mathematically, these synergies can be identified with factorization techniques

(Tresch et al. 2006), such as non-negative matrix factorization (NNMF) (Lee

and Seung 1999; Ting and Macpherson 2005; Tresch et al. 2006), principal

component analysis (PCA) (Krishnamoorthy et al. 2003; Krishnamoorthy et al.

2004) or independent component analysis (ICA) (Hart and Giszter 2004).

This modular organization of muscle activity has been demonstrated for

several motor behaviours in many species (Cappellini et al. 2006; d'Avella et al.

2006; Drew et al. 2008; Flash and Hochner 2005; Hart and Giszter 2004a; Hart

and Giszter 2004b; Krishnamoorthy et al. 2004; Krouchev et al. 2006; Latash et

al. 2005; Ting and Macpherson 2005; Torres-Oviedo et al. 2006; Torres-Oviedo

and Ting 2007; Yakovenko et al. 2011). Furthermore, the muscle synergies

extracted for one task frequently generalize and accurately reconstruct the

muscle activity patterns in another task, as has been shown for frog scratching,

jumping and locomotion (Cheung et al. 2005), human locomotion and running

(Cappellini et al. 2006) and forwards and backwards cycling motions (Ting et

al. 1999). Typically, a group of synergies is shared across tasks, although some

task-specific synergies may emerge according to the task (Chvatal et al. 2011;

Torres-Oviedo and Ting 2010). Also, the overall recruitment of the synergies

may change (Safavynia et al. 2011).

2.3 Mechanisms of postural control

2.3.1 Overview

Humans encounter numerous predictable and unpredictable

perturbations of balance during their everyday lives. In order to maintain

33

balance in these situations, the CNS relies on feedback and feedforward control

mechanisms to counteract any internal or external forces that may potentially

destabilize the body (Horak and Macpherson 1996; Massion 1992). In the case

of an externally driven perturbation of balance, such as an unexpected slip

when walking on a slippery surface, the CNS relies mainly on feedback

mechanisms to restore balance. In these situations, sensory input from a number

of sensors that signal the direction and amplitude of the perturbation is

integrated by the CNS to shape an appropriate automatic postural response that

effectively accelerates the position of the CoM to a position of stability within

the BoS (Henry et al. 2001; Horak and Macpherson 1996; Macpherson 1988a;

b; Ting and Macpherson 2004). In contrast, voluntary movements are

associated with disturbances to equilibrium due to internal forces related to the

muscle action for limb displacements. Here, the CNS is able to predict the

destabilizing effects of the interaction torques associated with the goal-directed

displacement of limb segments and programs appropriate postural adjustments

in anticipation of the disturbance (Bouisset and Zattara 1981; 1987; Massion

1992). These feedforward postural adjustments are commonly referred to as

'anticipatory postural adjustments' (APAs) and are programmed in feedforward

to compensate for the internally derived disturbance of equilibrium (Bouisset

and Zattara 1981; 1987; Massion 1992).

2.3.2 Intrinsic mechanical properties for stability

Several biomechanical and neural mechanisms are involved in

maintaining a desired postural configuration and equilibrium of the body. The

passive properties of the muscles and joints and the surrounding connective

tissues, as well as tonic muscle activity provide substantial contributions to

postural control (Horak and Macpherson 1996). For example, in upright quiet

stance, the alignment of the limb segments causes the primary muscles for

postural control, such as soleus and illiopsoas, to be stretched beyond their

resting length causing them to be tonically activated (Joseph and Nightingale

1952). When muscles are tonically active, there is an increase in the steepness

34

and decrease in the threshold of their length-tension curve enabling the muscle

to respond almost instantaneously against a disturbing force before any

peripheral or centrally-driven signals are available (Horak and Macpherson

1996). The passive properties of the muscles and their tonic activity provide

important load-compensating mechanisms not only for the maintenance of

posture, but also for providing stability during the execution of movement.

However, for larger, dynamic perturbations of balance, the viscoelastic

properties of muscles, ligaments and tendons, and muscle stretch alone are not

sufficient for restoring balance (Diener et al. 1984a). Thus, additional active

central mechanisms must be recruited for controlling equilibrium and

orientation when faced with unexpected and expected disturbances of balance.

These will be explored next.

2.3.3 Feedback postural responses

While standing on a moving bus, a person may experience a loss of

balance due to an unexpected acceleration or deceleration of the bus. In this

situation, the CNS responds quickly with stereotypical muscle activity in the

supporting limbs to oppose the disturbing force and restore equilibrium.

Postural responses of this nature rely on feedback from a weighted summation

(Mergner and Rosemeier 1998) of the sensory input from the available sensory

modalities, including visual, vestibular, proprioceptive, and somatosensory

systems, to detect the nature of the perturbation (Horak and Macpherson 1996).

These APR have been studied extensively in the laboratory setting and have

provided considerable information about the underlying neural control of

posture (Carpenter et al. 1999; Horak and Nashner 1986; Macpherson 1988a; b;

Nashner 1976; 1977; Torres-Oviedo and Ting 2007; 2010).

A spectrum of postural strategies for restoring balance following a

perturbation have been described in terms of the kinematic, kinetic and

electromyographic (EMG) patterns of the response (Do et al. 1982; Horak and

Macpherson 1996; Horak and Nashner 1986; Maki and McIlroy 1997; McIlroy

and Maki 1993a). For example, an unexpected backward translation of the

35

support surface induces a forward fall of the body. To restore equilibrium, the

CNS sequentially activates the extensor muscles of the body, namely, the

soleus, gastrocnemius, biceps femoris, the paraspinals and muscles of the neck,

to create the necessary GRFs that oppose the forward acceleration of the CoM

and reestablish an upright configuration of the body (Horak and Nashner 1986).

When the restoring torques are mainly focused at the ankle joint, the strategy is

known as an ankle strategy (Horak and Macpherson 1996; Horak and Nashner

1986). Alternatively, a hip strategy may be used to restore balance when the

perturbation is rapid, large in amplitude, or if the support surface is short

relative to the foot length (Horak and Nashner 1986). In this case, rotation at

the hips is accompanied by counter rotation of the ankle and neck to restore a

stable position of the CoM with respect to the BoS. In reality, however, most

foot-in-place postural responses typically involve a mixed strategy with

contributions from both the ankle and hip strategies (Horak and Macpherson

1996; Horak and Nashner 1986). Furthermore, when subjects are not

constrained to maintain their feet in place, an alternative strategy for regaining

balance may be to increase the BoS by taking a step (Do et al. 1982; Do et al.

1999; McIlroy and Maki 1993a; b) or using their hand to hold a support, such

as a handrail or wall (Maki and McIlroy 1997). The observation of these

different strategies led to several questions related to the nature of these

postural strategies: Are these complex movement patterns central programs?

How does the CNS select a particular strategy? Do these strategies reflect a

means by which the CNS simplifies the control of movement?

To gain insight about how the CNS organizes postural responses,

experimental paradigms have been designed where the support surface upon

which a subject is standing is unexpectedly translated or rotated to simulate a

slip. These perturbations elicit stereotypical activity in the muscles of the

supporting limb, typically evoked at latencies of 70-100 ms in humans (Diener

et al. 1988; Horak and Macpherson 1996; Horak and Nashner 1986; Nashner

1977) and 40-60 ms in cats (Horak and Macpherson 1996; Rushmer et al. 1988;

Ting and Macpherson 2004) following the onset of the disturbance. Using

36

perturbations in multiple directions of varying amplitudes and speeds, it has

been shown that the selection of a particular strategy, and the characteristics of

the APR, are shaped by the biomechanics (direction and velocity) of the

disturbance (Diener et al. 1984a; Macpherson 1988b; Ting and Macpherson

2004), as well as the initial body position (Horak and Moore 1993), prior

experience of disturbances (Macpherson 1994) and ‘central set’ (Horak et al.

1989a), which refers to the modulation of a response by descending commands

related to the expectation of what the perturbation and task parameters will be

(Brooks 1984; Evarts 1975; Schmidt 1982).

In response to unexpected perturbations of the support surface, both

humans and cats respond with muscle activity in the supporting limbs that is

broadly tuned to the direction of translation, with maximal amplitudes of

activation occurring for a single direction of translation (Henry et al. 2001;

1998b; Macpherson 1988b; Rushmer et al. 1988) or rotation (Carpenter et al.

1999; 2001; Ting and Macpherson 2004). While most muscles are characterized

by a unique curve, some similarities in the shape and tuning of multiple

muscles can be observed (e.g. (Henry et al. 1998b; Macpherson 1988b; Torres-

Oviedo and Ting 2007), suggesting that these muscles may be controlled as a

unit. Also, muscle activity is broadly tuned and shows some variability within a

reaching direction, indicating there is not a simple mapping between

perturbation direction and muscle activity (Moore et al. 1988). Together, these

findings suggest that the CNS may rely on a synergic organization to simplify

the task of selecting muscle activity patterns appropriate for a perturbation.

Recently, the muscle activation patterns for feedback postural control in

cats and humans have been examined in detail using factorization techniques

(Ting and Macpherson 2005; Torres-Oviedo et al. 2006; Torres-Oviedo and

Ting 2007; 2010). These studies have shown that a small number of spatially-

fixed muscle synergies are sufficient to reconstruct the trial-by-trial variability

of the complex spatial and temporal patterns underlying the postural responses

to perturbations in multiple directions (Ting and Macpherson 2005; Torres-

Oviedo et al. 2006; Torres-Oviedo and Ting 2007). Examination of the

37

relationship between muscle synergy recruitment and GRF patterns has also

shown that muscle synergies produce consistent end-point force patterns

(Krishnamoorthy et al. 2003; Ting and Macpherson 2005; Torres-Oviedo et al.

2006) and are stable in a variety of biomechanical contexts (Torres-Oviedo and

Ting 2010). More recently, shared muscle synergies have been shown for force

control in both non-stepping and stepping reactive postural responses (Chvatal

et al. 2011). Together, these results support the idea that muscle synergies

represent a neural mechanism for translating task-level goal into appropriate

muscle activation patterns for the task (Ting and Mckay 2007).

Characterization of the corresponding GRF for restoring equilibrium has

revealed that both cats and humans only produce force at the ground in one of

two directions aligned with the diagonal irrespective of perturbation direction

(Henry et al. 1998a; 2001; Macpherson 1988a). This strategy has been termed

the force constraint strategy (Macpherson 1988a) and is independent of prior

experience (Macpherson 1994), but is modulated by stance width (Henry et al.

2001; Macpherson 1994a). Specifically, net GRFs appropriate for opposing a

perturbation are achieved by modulating the amplitude, and not the direction, of

the force vector under each limb (Macpherson 1988a). In general, the force

constraint strategy is thought to simplify the control of posture by reducing the

number of parameters that the CNS must control (Macpherson 1988a).

2.3.4 Feedforward postural adjustments

Successful execution of voluntary movement during standing requires

postural activity to stabilize the disturbance of the CoM and compensate for the

disturbing interaction torques associated with the displacement of the focal limb

segments (Massion 1992). These adjustments of posture are termed

‘anticipatory postural adjustments’ (APA) since the postural changes occur

prior to, or at the same time, as the onset of the internal disturbance to posture

due to a voluntary movement and are programmed with a feedforward central

command (Massion 1992). It is traditionally believed that these feedforward

adjustments occurring prior to and during the movement are programmed

38

before feedback from the ongoing movement can influence them (Gahery 1987;

Massion 1992). Preparatory anticipatory postural adjustments (pPAs) occur

before the movement to assist in movement initiation or stabilization, whereas

associated anticipatory postural adjustments (aPAs) stabilize posture against the

internal reactive torques due to the limb movements (Bouisset and Zattara

1981; 1987; Commissaris et al. 2001; Cordo and Nashner 1982; Friedli et al.

1988; Friedli et al. 1984; Lee et al. 1987; Stapley et al. 1998; Stapley et al.

1999). Preparatory and associated postural adjustments will be referred to as

pPA and aPA, respectively, throughout this thesis.

Classically, APAs have been investigated during single arm raises

performed while standing. These studies have shown stereotypical activation

and inhibition of muscles in the supporting limbs, organized in a distal-to-

proximal sequence, occurs some 50-100 ms prior to the activation of the prime

mover (Belenkii et al. 1967; Bouisset and Zattara 1981; 1987; Marsden et al.

1977; Massion 1992). Biomechanical analyses of these movements have

demonstrated that the muscular activity in the legs is correlated to an upward

and forward acceleration of the CoM prior to the onset of the arm acceleration,

suggesting that APAs function to oppose the backward acceleration of the CoM

due to the arm raise itself (Bouisset and Zattara 1981; 1987). These findings led

to the traditional view that APAs occurring before the focal movement serve to

stabilize the position of the CoM in anticipation of the focal arm displacement.

This view is supported by extensive study of arm flexion (Belenkii et al. 1967;

Bouisset and Zattara 1987; Clement et al. 1984; Friedli et al. 1984; Horak et al.

1984; Lee et al. 1987), push or pull movements on a handle (Brown and Frank

1987; Cordo and Nashner 1982), and unilateral leg flexion tasks (Mouchnino et

al. 1992; Rogers and Pai 1990).

Anticipatory postural adjustments have also been attributed a role in

creating the dynamic forces necessary for initiating movement (Stapley et al.

1999). During locomotion, for example, preparatory muscle activity in the legs

serves to accelerate the CoM for gait initiation (Brenière and Do 1987; Brenière

and Do 1986; Lepers and Brenière 1995). Furthermore, the magnitude and

39

duration of the anticipatory EMG is correlated to the velocity of the first step

(Lepers and Brenière 1995), suggesting that the CNS predicts the dynamics of

the focal movement and integrates this information in planning the APA

according to the task constraints. Interestingly, this hypothesis is supported by a

number of studies in humans that have shown that APAs create the dynamic

forces to assist in the focal movement performance even when the BoS remains

fixed (Commissaris et al. 2001; Lee et al. 1987; Lee et al. 1990; Stapley et al.

1998; Stapley et al. 1999), suggesting that the role of APAs is not strictly for

postural stabilization. For example, when subjects are asked to lift an object or

push/pull on a handle during stance, activations across the postural muscles

serve to direct the GRFs to create the angular momentum that facilitates the

task(Commissaris et al. 2001; Lee et al. 1990). Thus, the view that APAs serve

to strictly stabilize the CoM does not necessarily generalize when a broader

range of movements is considered. Rather, it appears that APAs are

programmed as a function of the task requirements, serving to assist the

movement goals, be it stabilization or acceleration of the CoM.

More recently, substantial evidence in humans has accumulated

demonstrating that APAs are shaped by the biomechanics of a focal movement.

For example, in a reaction time paradigm, the temporal organization of the

APAs preceding single leg flexion is tightly linked to the time constraints of a

the focal task (Rogers and Pai 1990). Furthermore, during voluntary reaching or

pointing movements, APAs are programmed as a function of object distance

(Kaminski and Simpkins 2001; Stapley et al. 1998; Stapley et al. 1999), object

size (Bonnetblanc et al. 2004) and movement velocity (Bouisset et al. 2000;

Horak et al. 1984; Lee et al. 1987). Together, these results suggest that the CNS

anticipates and integrates the stability and dynamic requirements of the focal

movement into the postural command (Massion 1992). Thus, using knowledge

of the body and limb dynamics that is gained with experience and learning, the

nervous system is able to predict the dynamic consequences of the intended

movement and shape the postural adjustment appropriately (Bouisset and

Zattara 1981; 1987).

40

While the temporal organization of APAs in relation to the focal

movement has been studied extensively, less is known about the spatial

organization of feedforward postural activity in relation to voluntary

movements executed in multiple directions. Most studies investigating the

spatial organization of APAs have been restricted to focal movements in the

anterior-posterior direction only (Belenkii et al. 1967; Benvenuti et al. 1997;

Bouisset and Zattara 1987; De Wolf et al. 1998; Friedli et al. 1988; Friedli et al.

1984; Gantchev and Dimitrova 1996; Massion et al. 1999; Shiratori and Latash

2001; Slijper et al. 2002). These have shown that postural muscle activity is

sensitive to the direction of the focal movement and tends to be grouped

according to function. For example, groups of either frontal or dorsal muscles

are recruited for posterior and anterior movements, respectively (Bouisset and

Zattara 1987; Cordo and Nashner 1982; Friedli et al. 1984). However, a

detailed characterization of the directional sensitivity of the APAs, and how

muscles may be recruited synergistically is lacking.

To date, only a few studies have characterized the spatial organization

of APAs across a wider range of movement directions (Aruin and Latash 1995;

Santos and Aruin 2008; Vernazza et al. 1996). In particular, Aruin and Latash

(1995a) characterized the APAs preceding bilateral arm raises, performed as

quickly as possible across multiple directions, ranging from shoulder flexion

(forward) to shoulder extension (backward). Notably, they found that postural

muscles were preferentially activated for either forward or backward arm raises.

Differences in the tuning of the proximal and distal muscles were observed,

suggesting the recruitment of groups of functionally-related muscles. However,

in this study, movements were performed bilaterally and disturbances to the

CoM only occurred along the sagittal plane. How the CNS deals with planned

disturbances along the frontal plane as a result of asymmetrical reaching

movements remains to be examined.

The influence of directionality of an expected perturbation on the

organization of APAs was investigated by Santos and Aruin (2008). Here,

subjects were asked to intercept a pendulum with their right or left hand

41

released by the experimenter from a fixed distance and height. In order to test

the effect of directionality, subjects were positioned so as to produce

perturbations in either the frontal, sagittal or oblique planes. The authors

characterized the APAs in muscles distributed across the body, with an

emphasis on the lateral muscles of the body. In general, their findings extend

those of Aruin and Latash (1995a) by demonstrating the directional sensitivity

of APAs holds for the lateral muscles of the body. Furthermore, they show that

complex patterns of muscle coordination are required for adjusting posture in

preparation for intercepting the target, with combinations of muscles acting

together. The authors suggest that the CNS relies on the flexible recruitment of

muscle synergies to produce task appropriate modulation of muscle activity.

Together, the studies of Aruin (1995; 2008) show distinct patterns of

feedforward postural activity that are sensitive to the direction of the impending

movement. Whether these patterns for controlling posture and balance are

similar to those observed for feedback postural control which are triggered by

afferent feedback, is not known.

2.4 Models for movement control

Humans are able to execute complex movement patterns quickly and

with exceptional accuracy, while interacting with objects in their environment

whose dynamic characteristics may or may not be known. In addition, the CNS

must also consider the inherent delays associated with biological feedback and

the nonlinear relationship between muscle and force generation. From a

computational view, a fundamental question is what models are used by the

CNS to compute the appropriate muscle activation patterns that will results in

efficient movements appropriate for the task and environmental context

(Desmurget and Grafton 2000)?

42

2.4.1 Goal-directed movements require both feedback and feedforward

control mechanisms

Given the rapidity and smoothness of human motion, it is widely

accepted that the CNS cannot rely exclusively on feedback mechanisms to

coordinate movements (Kawato 1999). Similarly, a purely feedforward, or

'predictive,' model programmed entirely before the movement onset would also

be limited in it's ability to respond to unexpected perturbations or errors

occurring during the movement. Consequently, it is proposed that the CNS

relies on a 'hybrid' of feedforward and feedback control mechanisms (Gritsenko

et al. 2009), or a dual-mode of control (Milner 1992), for planning and

executing complex goal-directed movements (Desmurget and Grafton 2000).

Early formulations of this model estimated that the CNS programmed the initial

trajectory in a feedforward manner and only accessed sensory feedback in the

late stages of the movement to ensure accuracy to the target goal (Milner 1992).

More recently, these ideas have been developed in terms of internal models, or

representations, that the CNS has of its own dynamics and the external

environment. The CNS organizes an approximate motor plan prior to the

movement, which is then executed under the 'supervision' of internal feedback

loops that may have non-sensory contributions, which can then modulate the

original plan in real time (Desmurget and Grafton 2000).

Historically, a feedforward model for controlling voluntary movements

had received favorable review in the literature on account of the sensorimotor

delays documented for feedback control were too long to effectively control

movement. For example, sensory feedback from the vestibular or visual system

requires a minimum of 80-100 ms but may be as long as 300-700 ms before it

can influence an ongoing movement, such as visually-guided reaching

(Desmurget and Grafton 2000; Jeannerod 1988; Paillard 1996). However, given

that the CNS is able to respond to errors and modify a motor program online, it

is suggested that feedback signals must be accessible and integrated by the

CNS. Also, movements are more accurate when sensory feedback is present,

highlighting the contributions of feedback to the overall control of movements.

43

Current models stress the importance of strong internal feedback loops, which

rely on internal estimates of efferent and afferent signals, as the primary error-

detecting signal (Davidson and Wolpert 2005; Desmurget and Grafton 2000;

Wolpert and Miall 1996).

2.4.2 Internal models

An important insight that has led to current hypotheses about how the

CNS effectively deals with the inherent delays associated with feedback control

and the nature of the feedback signal was postulated by von Holst and

Mittelsteadt (1950). They proposed that the CNS is able to distinguish between

self-induced motion and a passive displacement of the body with respect to the

environment by storing a 'copy' of the motor command for the goal-directed

motion within the brain (Desmurget and Grafton 2000). An extension of this

idea is that the CNS has knowledge of the properties of the body and its

surrounding environment (Davidson and Wolpert 2005; Kawato 1999; Scott

and Norman 2003). Therefore, the CNS can make predictions about the

expected sensory consequences resulting from internally generated movements

and drive corrections when discord between the expected and observed sensory

signals occurs (Davidson and Wolpert 2005; Wolpert and Miall 1996). These

ideas form the basis of the concept of internal models.

Internal models are defined as the neural processes that estimate the

input/output relationships between the motor system and the CNS (Kawato

1999). Two types of internal models have been proposed. A forward model

estimates the sensory consequence of internally driven motion of the body

using an efference copy of the motor command (Wolpert and Miall 1996). In

contrast, an inverse model employs the reverse transformation to generate the

neural commands for movement based on the desired outcomes or

consequences of that movement (Wolpert and Miall 1996). The internal model

hypothesis predicts that the CNS acquires an inverse model of the body's

dynamics through motor learning, enabling the control of movements in a

feedforward manner (Kawato 1999). In turn, a forward model of an efferent-

44

based copy of the future limb state allows for rapid correction of movements

via feedback mechanisms (Davidson and Wolpert 2005; Kawato 1999). The

existence of internal models is not restricted to the motor system and has been

used extensively to explain processes of the sensory system also (Hess and

Angelaki 1999; Merfeld et al. 1999; Roy and Cullen 2004).

2.4.2.1 Inverse internal models

When performing a voluntary movement, such as a reaching movement

to a target, the CNS calculates the motor commands to generate muscle activity

that will compensate for the arm dynamics and result in the desired kinematic

trajectory of the arm (Kawato 1999). However, in situations where the arm is

faced with novel dynamics, the motor program is no longer appropriate and an

erroneous trajectory is produced. To examine this process, subjects have been

studied while performing reaching movements in dynamic force fields that alter

the dynamics of the environment (Lackner and Dizio 1994; Shadmehr and

Mussa-Ivaldi 1994; Shadmehr et al. 1993). In these paradigms, the force fields

generate consistent forces that depend on the state space, such as the position or

velocity of the segment, and cause deviations of the arm trajectory. Over

repeated trials, subjects learn to perform straight trajectories by internalizing the

dynamics of the environment and modifying their muscle activity (Bhushan and

Shadmehr 1999; Flanders 2011; Ioffe et al. 2007). However, when the force

field is unexpectedly removed, a large error is produced by the subject,

providing evidence that subjects have adapted their model of the inverse

dynamics of the arm (Kawato 1999; Shadmehr and Krakauer 2008).

2.4.2.2 Forward internal models

In order to compensate for the delays inherent to sensorimotor

processes, it has been postulated that the execution of skilled voluntary

movements relies on a prediction of the future state of the motor system using

an efferent-based copy of the motor command (Davidson and Wolpert 2005).

This allows for rapid, corrective responses during movement executions

45

(Gritsenko et al. 2009). Evidence from studies examining how grip force on an

object is adjusted in relation to predictable and unpredictable changes in load

force have substantiated the use of forward models by the motor system

(Davidson and Wolpert 2005; Flanagan and Wing 1997; Johansson and Cole

1992; Kawato 1999). For example, when using a precision grip to hold an

object with stable properties between the fingertips and thumb, the CNS

predicts the changes in load force caused by any self-induced changes in load

force due to voluntary movement of the limb and adjusts grip force in parallel

with the change in load force to prevent slippage of the object (Johansson and

Westling 1984). However, if the object's dynamics are unstable or not known,

grip force is adjusted in a reactive manner to changes in load force. These

results suggest that the CNS uses predictive mechanisms, such as a forward

model, to prevent unwanted delays in motor execution and ensure stability of

movements. Similarly, such predictive mechanisms have been documented for

maintaining the equilibrium of the CoM during the execution of goal-directed

arm movements (Bouisset and Zattara 1981; 1987). Recently, such

computational frameworks have predicted that the CNS relies on state

estimation based on an internal model of the body and sensor dynamics to

process afferent information and calculate the orientation of the body (Kuo

2005).

2.5 The control of voluntary arm movements

Goal-directed reaching movements have been studied extensively in

humans and primates to understand how the CNS plans movements and

integrates afferent and efferent signals into the control mechanism (see review,

(Georgopoulos 1986). The kinematic, dynamic and muscle activity for visually-

guided reaching movements are well documented, although the roles of

different sensory inputs to the planning and execution of the reach remain

highly debated (Sarlegna and Sainburg 2009). It appears that reaching

movements to visual targets are planned using information about the initial

position of the hand and the final target position. Likely, initial hand position is

46

encoded by a weighted summation of proprioceptive and visual feedback,

whereas the end of reach position is highly influenced by visual information of

the target (Rossetti et al. 1995). Following illumination of the target, there is a

sequential organization of the eyes, head and arm toward the target (Biguer et

al. 1982). Furthermore, the arm trajectory to the target is characterized by a

bell-shaped velocity curve. Typically, the acceleration and deceleration phases

are of equal duration, however the deceleration phase may be lengthened when

vision of the arm or hand position is restricted (Elliott et al. 1991) or when

increased accuracy is required, such as when reaching to targets reduced in size

(Soechting 1984).

2.5.1 Online control of visually-guided reaching movements

The contributions of the visual and proprioception systems to the

control of aimed reaching movements have been examined more closely in

paradigms where an unexpected shift in target position is introduced. These

double-step paradigms, initially developed to study saccadic eye movements

(Becker and Jürgens 1979; Levy-Schoen and Blanc-Garin 1974; Wheeless et al.

1966), have provided a means to investigate the underlying visuomotor

processes that contribute to the rapid, smooth correction of the arm trajectory

during visual perturbations of reaching movements. When a shift in target

position occurs very early in the reach trajectory, such as immediately before or

at reach onset, very short latency corrections of the reach occur (Day and

Brown 2001; Day and Lyon 2000; Pélisson et al. 1986; Prablanc and Martin

1992). Subjects successfully correct their arm trajectory, as indicated by a lack

of inflection point in the kinematic profile of their reach, at latencies similar to

reaction times for unperturbed reaches without significantly increasing their

movement time. These online corrections are presumed to be automatic in

nature, given that they occur in situations where the subjects are unaware of the

target shift, such as during the saccadic suppression (Bard et al. 1999; Johnson

and Haggard 2005; Pélisson et al. 1986). However, when the target shift occurs

very late in the deceleration phase of the movement, subjects were unable to

47

successfully correct their reaching movement (Komilis et al. 1993). Vision of

the arm is not critical for making correction, however end-point accuracy of the

reaching movement is improved when subjects can visualize their arm and the

target (Komilis et al. 1993; Pélisson et al. 1986; Sarlegna et al. 2003).

2.5.2 Standing imposes equilibrium constraints during perturbed reaching

An interesting question that arises from the literature documenting the

online control of voluntary movement is, how does the CNS resolve

equilibrium constraints when corrective reaches are performed in the standing

position? Answering this question is fundamental to understanding why falls

often occur in dynamic situations where both the movement and postural

systems are challenged. When arm movements are perturbed in the standing

position, the CNS must correct the perturbed arm movement in addition to

dealing with the equilibrium constraints imposed by the upright posture and

decreased BoS.

To date, only two studies have investigated how the CNS corrects goal-

directed movements when standing (Fautrelle et al. 2010; Martin et al. 2000).

Using a modified version of the classic double-step paradigm, Martin and

colleagues (2000) investigated how the likelihood of an impending shift in

target position along the sagittal plane affected the kinematics for reaching

movements executed in the standing position. They found that the kinematics of

the arm movement remained stable across conditions, however trunk flexion

was increased to compensate for the uncertainty of the reaching movement,

suggesting that uncertainty about the movement was integrated in the motor

program for posture and movement in a predictive fashion (Martin et al. 2000).

Similarly, Fautrelle and colleagues (2010) examined the modulation of

postural control during online corrections of reaching movements in standing.

They showed that rapid corrections for arm and leg muscles occurred at

approximately the same latency of 100 ms. At such short latencies, the authors

suggested that the online corrections of posture and movement occur

automatically and are organized at subcortical levels through rapid internal

48

feedback loop mechanisms (Fautrelle et al. 2010). It is not clear from their data,

however, whether the correction in the leg preceded that of the arm. More

detailed information about the temporal organization of the online corrections

in posture and movement will provide additional insight into the central

organization of posture and movement. Therefore, it remains to be seen whether

the CNS updates the commands for posture predicatively of the impending arm

correction, or whether the CNS relies on information from the arm movement

to update posture for the maintenance of balance.

2.6 Summary and direction for future investigation

Several important questions related to the nature and central

organization of postural control during voluntary movements remain to be

addressed. Most studies investigating the central control of these two

behaviours have focused on the temporal coupling between the ‘GO’ signal

triggering a voluntary movement, the onset of the postural changes

accompanying the movement and the onset of the movement related changes

themselves. While the studies detailed in this chapter have provided

fundamental insight into the coordination of feedforward postural adjustments,

a detailed examination of the spatial organization of pPAs and aPAs is lacking.

This information will allow for comparison to other types of postural

behaviour, namely the strategies for maintaining balance organized via a

feedback mode of control. It is expected that this knowledge will contribute to

elucidating the neural structures involved in organizing the general strategies

for posture control.

A second important question is how the CNS resolves the constraints for

equilibrium control when voluntary movements executed in standing are

perturbed and online corrections in the movement are necessary. Specifically,

how are commands for posture and movement controlled in relation to one

another? What models are adopted for ensuring successful execution of the

task? It is commonly accepted that correcting movement errors in the standing

position presents a dynamic and challenging situation for many, as evidenced

49

by the frequency of falls occurring in these types of situations. Therefore, it is

important to investigate the underlying mechanisms for posture and movement

control in order to prevent falls and rehabilitate fall-prone individuals.

The body of work presented here aims to further our understanding the

mechanisms of equilibrium control by examining the strategies for maintaining

balance during goal-directed reaching movements to targets located in multiple

directions by focusing on the above-mentioned topics.

50

51

Chapter 3

General Methodology

The general aim of the thesis was to gain insight into how the CNS

coordinates the control of posture and voluntary movement. To investigate the

questions central to this thesis, an experimental protocol was designed where

subjects were required to reach towards targets in multiple directions while

standing. The rational for adopting this paradigm and the general characteristics

of the experimental paradigm common to each study will be presented here.

The specific experimental details for each are provided in Chapters 4, 5 and 6,

respectively.

3.1 Rational for experimental protocol

Traditionally, questions related to the coordination of posture and

movement have been investigated by characterizing the kinematics, kinetics and

muscle activity generated during arm raising (flexion/abduction) movements

while standing (Bouisset and Zattara 1981; 1987).While this classical approach

allows for a clear distinction between postural and focal components of the

movement, it is limited in it’s ability to fully explore the spatial organization of

APAs, given that the arm motion in these paradigms is generally limited to only

a few directions. Furthermore, it has been previously highlighted that in order

to assess the structure of the data set and determine whether muscle synergies

reflect the organization of the CNS and not the structure of the experimental

design (Macpherson 1991), the experimental paradigm should have more

muscles and experimental conditions than the number of synergies that could be

expected (Ting and Mckay 2007). Thus, in our experiments, to examine the

spatial and synergic organization of feedforward postural adjustments

preceding and accompanying reaching movements, directionality of postural

52

activity was established by requiring subjects to point to multiple targets (a total

of 13), interspersed over a range of 180°.

A primary objective of the thesis is to explore the spatial organization of

feedforward postural activity in the context of what is known about feedback

postural control to generate knowledge about the central control of posture and

movement. This will be addressed in Chapters 4 (SA1) and 5 (SA2). In

response to an unexpected perturbation of balance, APRs accelerate the CoM

and body to a region of stability within a fixed BoS (Horak and Macpherson

1996). These feedback responses depend mainly on afferent signals (Allum et

al. 1994; 1993; Stapley et al. 2002) that shape the timing and amplitude of the

response. In contrast, feedforward postural commands are centrally generated

with descending commands to postural muscles (i.e., rather than being triggered

by an afferent signal) (Massion 1992). In order to make comparisons between

the organization of these modes of control, a paradigm is needed in which the

task goals are similar: to accelerate the CoM and body upon a fixed BoS. In this

regard, a task requiring feedforward postural activity to accelerate the CoM

within a fixed BoS would more closely match the behavioural goals of APRs,

enabling more direct comparisons between feedback and feedforward modes of

postural control. Thus, a reaching task (described below in detail) was designed

where subjects performed reaching movements in standing to targets located

just beyond arm’s length in multiple directions. This paradigm provides an

excellent opportunity therefore, to characterize and quantify feedforward

postural activity in a task where the goal is to accelerate the CoM upon a fixed

BoS.

The third objective (SA3) of the thesis is to examine the nature of the

control signal underlying aPAs accompanying voluntary movements in

situations where the task goal changes after the focal movement has been

initiated. Previous attempts to explore the online control of posture during

voluntary movement have utilized a visual double-step paradigm. However,

these studies have focused mainly on how uncertainty about target location

affects movement performance during standing (Martin et al. 2000) or how

53

quickly the motor system can initiate a correction to the arm movement given

the additional equilibrium constraints of standing (Fautrelle et al. 2010).

Although these studies provide important insight about movement corrections

in standing, neither specifically addressed how posture is updated with respect

to the movement correction. It is important to know how the CNS achieves this

control, and to determine whether the movement correction relies on a postural

correction for successful execution of the movement as people often experience

falls in dynamic situations (Inglis and Macpherson 1988). Thus, borrowing

from the double-step paradigm, we were able to specifically quantify the

postural adjustments occurring just before, and concurrently with the correction

in arm trajectory so as to provide insight regarding the interaction of posture

and movement in these dynamic motor tasks.

3.2 Overview of experimental protocol

3.2.1 Experimental apparatus

The general experimental set-up utilized for SA1 and SA2 is shown in

Figure 3.1. For all experiments, right-handed subjects performed reaching

movements with their dominant arm to targets located in multiple directions

while standing. Subjects stood barefoot on two triaxial force plates (model

FP4060; Bertec, Columbus, OH), which were centred in a custom-built target

light array, fully adjustable in height and target distance. The light array was

suspended from the ceiling and could not provide support to the subject upon

contact with the target. The array contained 13 light-emitting diodes (LED)

targets, interspersed by 15°, with the 0° target located to the subject’s right, 90°

at their midline and 180° to their complete left. Each target consisted of a red

LED encased in a modified gaming switch (model 459512; RP Electronics,

Burnaby, BC, Canada) that produced a 5V pulse upon depression of the switch.

Targets were placed at shoulder height, at a standardized distance of 130% of

the subject’s reach length, as measured to each target. Pilot tests demonstrated

that this distance ensured that significant adjustments in posture were engaged

54

to attain the target, while maintaining a fixed BoS. Reach length was measured

between the subject’s xiphoid process (point from which they initiated their

pointing movements) and the tip of the right index finger when the arm was

extended to each target. Measurements were performed with the subject centred

in the array (xiphoid process aligned with the 90° target) while they stood

straight, and maintained neutral scapular retraction. For leftward targets,

rotation of the trunk was permitted in order to maintain the shoulders square to

the target. This was done to determine the distance to reach the target in

postural configurations that most similarly matched those adopted during the

reaching movements.

Figure 3. 1: General experimental set-up. Plan view of target light array

illustrating subject orientation on the force plate relative to the array. Targets

are arranged from right (0°) to left (180°).

55

3.2.2 Behavioural task

Questions related to the feedforward control of posture were explored

by quantifying the postural adjustments preceding and accompanying a discrete

reaching task performed in standing. For all experiments, subjects adopted their

natural stance width, with one foot on each of the adjacent force plates and their

midline aligned to the 90° target. Subjects began each trial with their left arm

hanging loosely by their side and their right index finger depressing a switch

affixed to their chest, which when released provided a 5V step signal. Upon

random illumination of one of the target LEDs, subjects were instructed to point

and depress the illuminated target at their preferred movement speed.

Immediately upon contact with the target, subjects were asked to release the

target but hold the body configuration with their arm outstretched to the target

for a period of 2 seconds (indicated by extinction of the illuminated LED). By

requiring subjects to stabilize their body at the target, it was possible to

examine whether task goal (acceleration versus stabilization of CoM) affected

the organization of the feedforward postural adjustments. No constraints with

respect to movement speed or finger trajectory were specified.

3.2.3 Protocol specific to SA1 and SA2

In SA1 (Chapter 4) and SA2 (Chapter 5), subjects performed discrete

reach to point movements to one of 13 targets. Testing began with an

acclimatization period where subjects pointed to each target twice, presented in

random order. Following these trials, subjects executed a total of 15 trials to

each target (195 trials) and 15 catch trials where no target illuminated. Catch

trials were presented in order to minimize prediction of the occurrence and

direction of the illuminated target. All target directions were presented in a

pseudorandomized order, in blocks of approximately 50 trials interspaced with

5-minute rest period to reduce the effects of fatigue. Details of the calculated

56

data variables and analysis for SA1 and SA2 are presented in Chapters 4 and 5,

respectively.

3.2.4 Protocol specific to SA3

A modified version of the classic double-step paradigm was developed

for investigating the nature of the control underlying associated postural

adjustments accompanying voluntary reaching movements (SA3; Chapter 6).

Specifically, the basic reaching paradigm of SA1 was modified so that a

voluntary reaching movement could be disrupted by changing the task goal

following onset of the reach. This was achieved by programming a shift in the

illuminated target after the onset of the reaching movement (release of the chest

switch). In these experiments, four targets were used: the central, or 90°, target,

and the three consecutive targets to the right of the central target (45°, 60°and

75°). Two types of trials were presented to subjects: (i) “reach” trials that

consisted of discrete reach movements to the central target (90°) and (ii) “corr”

trials, that required an online correction of the arm trajectory to one of the other

three targets after a reach was initiated to the 90° target. As in SA1, subjects

stood barefoot on two triaxial force plates, with their xiphoid process aligned to

the central target (90°). Each trial began with the subject depressing the chest

switch with their right index finger and their left arm hanging loosely relaxed

by their side. After a variable delay, the central target (90° or L1) would

illuminate. Subjects were instructed to reach and press the target at their natural

speed. Upon contact with the target, subjects were to release the target but

maintain the postural configuration with their arm outstretched to the target

until the target LED extinguished, which corresponded to a period of 2 seconds.

On some trials (corr trials), the target light would shift from L1 to any one of

the other three targets: 75° (corr75), 60° (corr60) or 45° (corr45) after a variable

delay from the reach onset. Subjects were instructed to point to the newly

illuminated target (L2) upon detection of the light shift. The target shift would

occur after a variable delay following online detection of the voltage drop of the

chest switch signal upon release of the chest switch. Subjects performed a block

57

of acclimatization trials consisting of five regular reaches to each target.

Subjects were subsequently presented with trials in random order, including at

least 60 reach trials to the 90° target, 15 trials to each of the corr conditions

(corr75, corr60 and corr45), and 15 catch trials were no target illuminated.

Subjects typically performed blocks of 40 trials, separated by rest periods of 5

minutes.

3.2.5 Data collection and analysis

For all experiments, ground reaction forces (GRF) and moments in the

mediolateral (x), anteroposterior (y) and vertical (z) axes were recorded with

two triaxial force plates (model FP4060; Bertec, Columbus, OH) sampled at

1000-Hz. Three-dimensional body kinematics were collected using a six camera

MX3 motion-capture system (Vicon Peak, Lake Forest, CA) that sampled at

200 Hz. Muscle activity from a total of 16 leg, trunk and arm muscles was

recorded using two DelSys Bagnoli 8-channel systems (Delsys, Boston, MA).

Custom-written programs written in LabVIEW (National Instruments, Austin,

TX) controlled the illumination of the target lights and acquired signals from

the chest and target switch. Synchronization with the analog signals from the

force plates and EMG system was done using the Vicon controller. Custom

programs were written in Matlab (The MathWorks, Natick, MA) for processing

and analyzing the collected data. The details of the data analysis for each

experiment are presented in the respective chapters.

3.3 Significance of the experimental paradigm provides basis for

further exploration

The experimental paradigm developed for this thesis provides a basis

for investigating fundamental questions related to the coordination of posture

and movement. In the first place, reaching tasks in standing provide a motor

task where there is a clear dissociation of postural and focal components of the

movement. Furthermore, by examining postural control strategies adopted

58

when performing reaching movements in multiple directions, we are able to

contribute knowledge about the spatial organization of feedforward postural

adjustments that complement current knowledge about feedback-based postural

responses to rotations and translations of the support surface. Finally, this

paradigm provides a unique opportunity to fully explore the nature of deficits of

postural control observed in populations with pathologies of the motor systems,

such as aging or Parkinson’s Disease. Comparisons of the temporal and spatial

patterns of postural muscle activity to baseline measures determined for healthy

individuals can provide important insight into the mechanisms of the disrupted

motor behaviour.

59

Chapter 4

Reaching to multiple targets when

standing: The spatial organization of

feedforward postural adjustments

4.1 PREFACE

The first study of this thesis was motivated by a perceived gap in

knowledge regarding the strategies for feedforward postural control for

reaching movements in standing. Previous studies of support surface

translations in multiple-directions have demonstrated the existence of a force

constraint strategy (Henry et al. 2001; Macpherson 1988a) and muscle activity

that is directionally-tuned (Henry et al. 2001; Macpherson 1988b) for feedback

postural control. Whether the CNS relies on similar strategies for simplifying

the control of posture organized in feedforward is not known. Therefore, as a

primary aim, this chapter examined the spatial and temporal organization of the

GRF and EMG patterns of feedforward postural adjustments proceeding and

associated with pointing movements in multiple directions executed in

standing. Furthermore, the experimental paradigm developed for the study in

this chapter provides a novel means for investigating the coordination of

posture with respect to voluntary movements.

This chapter was adapted from Leonard JA, Brown RH, and Stapley P.J.

Reaching to multiple targets when standing: The spatial organization of

feedforward postural adjustments. Journal of Neurophysiology 101(4): 2120-

2133, 2009. This manuscript has been reprinted with permission from The

American Physiological Society, publisher of the Journal of Neurophysiology.

The paper is presented in the same format in which it was published with the

exception of formatting to figures and tables to comply with McGill University

thesis formatting guidelines.

60

4.2 ABSTRACT

We examined the spatial organization of feedforward postural

adjustments produced prior to and during voluntary arm reaching movements

executed while standing. We wished to investigate if the activity of postural

muscles before and during reaching was directionally tuned and if a strategy of

horizontal force constraint could be observed. To this end, 8 human subjects

executed self-paced reach-to-point movements upon the random illumination of

1 of 13 light targets placed within a 180° array centered along the mid-line of

the body. Analysis was divided into 2 periods: a first corresponding to the 250

ms preceding the onset of the reaching movements (termed pPA period), and a

second 250 ms period immediately preceding target attainment (the aPA

period). For both periods, EMG activity of the lower limb muscles revealed a

clear directional tuning, with groups of muscles being activated for similar

directions of reach. Analysis of horizontal ground reaction forces supported the

existence of a force constraint strategy only for the pPA period, however, with

those in the aPA period being more widely dispersed. We suggest that the

strategy adopted for feedforward pPAs is one where the tuned muscle synergies

constrain the forces diagonally away from the centre of mass (CoM) to move it

within the support base. However, the need to control for final finger and body

position for each target during the aPA phase resulted in a distribution of

vectors across reaching directions. Overall, our results would support the idea

that end-point limb force during postural tasks depends upon the use of

functional muscle synergies, which are used to displace the CoM or decelerate

the body at the end of the reach.

4.3 INTRODUCTION

For multi-joint movements executed during standing such as reaching

forwards, postural adjustments occurring prior to movement onset shift the

CoM within the base of support in order to initiate the movement, and

61

associated postural adjustments overcome the postural disturbances related to

movements of the limbs (Bouisset and Zattara 1981; 1987). To ensure a

controlled transition from one postural configuration to another, these

adjustments of posture must be planned by the central nervous system (CNS) in

advance, and a feedforward mode of neural control sends commands to both

focal and postural muscles to initiate and stabilize posture. Both the preparatory

and associated postural adjustments (pPAs and aPAs, respectively) are

considered to be feedforward in nature as they are produced before feedback

from the ongoing movement can influence them (Gahery 1987; Massion 1992).

The objective of the present study was to examine the spatial organization of

postural muscles and forces produced before and during voluntary reaching

movements in multiple directions to gain insights into the nature of their

underlying control.

Feedback mediated postural responses to unexpected disturbances of

balance have been well characterized. When the surface upon which humans

are standing unexpectedly moves, the body is destabilized in the direction

opposite to that of the surface displacement. In order to regain balance, humans

produce short latency automatic postural responses (APRs) in the supporting

limbs that oppose the perturbation and drive the CoM back towards its initial

position relative to the support surface (Horak and Nashner 1986). The latency

from the initiation of the support surface movement to the onset of the evoked

EMG response is in the order of 80-120 ms in humans (Horak and Macpherson

1996; Nashner 1977; Ting and Macpherson 2004). These compensatory APRs

are triggered by somatosensory feedback from the feet and legs (Bloem et al.

2000; Bloem et al. 2002; Horak and Macpherson 1996; Stapley et al. 2002) and

unless prior warning of the upcoming perturbation is given (Jacobs and Horak

2007; McChesney et al. 1996), they are produced entirely using a feedback

mode of neural control.

Studies in animals and humans have examined feedback-based APRs to

unexpected translations of the support surface in multiple directions with the

aim of identifying strategies that the CNS may adopt to simplify the control of

62

perturbed stance (Fung et al. 1995; Henry et al. 2001; 1998b; Macpherson

1988a; b; Ting and Macpherson 2004). In the cat, regardless of the direction of

horizontal platform translation, force vectors were exerted in one of two main

directions at each limb (termed the ‘force constraint strategy’, (Macpherson

1988a). Muscle activity displayed a tuning across directions with maximal

amplitudes for each muscle arising for one specific direction (Chanaud and

Macpherson 1991; Macpherson 1988b). Similar results have also been obtained

in standing humans subjected to postural perturbations (Henry et al. 2001;

1998b). Overall, results have supported the hypothesis that force vector

production is a high level parameter adopted to reduce the multiple of degrees

of freedom associated with complex postural tasks, and that the production of

the desired vector is solved at a lower level by the synergic recruitment of

muscle groups. More recent studies in the standing cat have elaborated this

hypothesis, and have suggested that an internal model of postural force

generation coordinates functional muscle synergies rather than biomechanical

limb constraints alone (McKay et al. 2007; McKay and Ting 2008). However,

the production of a force constraint upon the relationship between the muscle

synergies produced and the current limb geometry (Torres-Oviedo et al. 2006).

Recent studies have shown that neural commands for feedforward

postural adjustments can be identified in the pontomedullary reticular formation

(PMRF) of the brainstem (Schepens and Drew 2006; 2004; Schepens et al.

2008). Neurons in this area discharged either during the pPA, the aPA or both

during reaching movements in the standing cat (Schepens and Drew 2006;

2004). The activity of this area has also, however, been shown to be implicated

in compensatory, or feedback-based, postural control. Microstimulation of

peripheral afferents, known to be essential for triggering short latency APRs

(Stapley et al. 2002), activate particular subsets of reticulospinal neurons (Drew

et al. 1996) and PMRF neurons are activated strongly during postural

perturbations in the standing cat (Stapley and Drew 2009). If specific neural

sites can modulate postural responses with different underlying modes of

control (i.e., feedforward or feedback), it is plausible to predict that the outward

63

expression of those postural adjustments may show similarities in their

organization. The present study aimed, therefore, to investigate if a similar

spatial organization of horizontal force and muscle activity to that seen with

feedback-based postural responses could be observed for predictive

feedforward postural adjustments in humans, despite their different modes of

neural control. To generate directionality, we asked human subjects to reach

and point to targets in multiple directions throughout 180° at shoulder height

whilst standing. Postural adjustments immediately preceding the onset of the

focal limb movement (pPAs) and those produced before target attainment

(aPAs) were recorded and analyzed. We hypothesized that: 1) feedforward

postural adjustments are directionally-tuned to the current goal of the task

(initiation or termination of the reaching movements), and 2) ground reaction

forces show directional force constraint, supporting the idea that such a strategy

is a high level parameter adopted by the CNS, regardless of the mode

(feedforward or feedback) of neural control.

4.4 MATERIALS AND METHODS

4.4.1 Subjects

Eight (1 female and 7 male) healthy subjects, without any known

neurological, visual or orthopedic disorders, were recruited from the McGill

University student population to participate in the present study. Subjects had a

mean age of 22.5 ± 4.3 (SD) years, a mean height of 1.74 ± 0.04 (SD) m, and a

mean weight of 66.6 ± 5.9 (SD) kg. They were all right-hand dominant. They

gave their informed consent to participate and experiments were conducted with

the approval of the McGill University research ethics board

4.4.2 Experimental apparatus and set-up

Subjects stood barefoot on 2 tri-axial force plates (model FP4060,

Bertec Corp., Columbus, OH, USA) that measured ground reaction force (GRF)

and moments in mediolateral (X), anteroposterior (Y) and vertical (Z) axes at

64

1000 Hz. Stance width (the mediolateral distance between the feet) was taken

as the average of each subject’s natural stance distance as measured after 3

trials of walking the length of the laboratory. Foot position was marked for each

subject and care was taken that subjects kept their feet in the same position for

the duration of the experiments. They were centered in a custom-built 180°

light target array, fully adjustable in height and target distance (see Fig. 1A).

The array contained a total of 13 light emitting diodes (LEDs) each spaced at

15° intervals. Light targets were 2.5 cm in diameter and consisted of 5V red

LEDs encased in modified gaming switches (model 459512, RP Electronics,

Burnaby, BC) that produced a 5V pulse upon contact. The gaming switches

were mounted at the ends of lightweight aluminum dowels adjustable in length,

affixed to a semi-circular aluminium bar suspended from the ceiling. A chest

band worn by all subjects was also equipped with the same switch in order to

detect movement onset (see below).

The activity of 16 muscles was recorded using 2 DelSys Bagnoli 8-

channel systems (Boston, MA, USA) at 1000 Hz. The following muscles were

recorded bilaterally (left and right legs): tibialis anterior (TAl and TAr,

respectively), soleus (Soll, Solr), lateral gastrocnemius (GasLl, GasLr),

peroneus longus (Perl, Perr), biceps femoris (BFl, BFr), rectus femoris (RFl,

RFr) and tensor fascia latae (TFLl, TFLr). In addition, anterior and posterior

deltoid was recorded at the right shoulder. Bilateral kinematic data were

collected using a 6 camera MX3 motion capture system (ViconPeak Inc., Lake

Forest, CA, USA) sampling at 200 Hz. A total of 36 markers were placed at

different locations on the subjects’ whole body as listed in the Plug-in-Gait

model (Vicon Peak ®). This model provides an accurate estimate of CoM

position when compared to the method of ground reaction force integration

(Gutierrez-Farewik et al. 2006). Analog signals from the force plates and EMG

system were captured through the Vicon MX3 controller. A customized

program written in LabView (National Instruments, Austin, TX, USA) was

used to control the illumination of target lights, acquire and synchronize digital

65

signals from the switches (target and chest), and initiate and synchronize data

collection with the Vicon system.

4.4.3 Experimental Procedures

Subjects were centered in the array with their mid-line (xiphoid process)

aligned with the 90° target direction and the 0° and 180° target directions

representing, respectively, each subject’s far right and left side targets (see Fig.

4.1A). Upon the random illumination of 1 of the 13 target lights, subjects were

asked to point and press the lighted switch. Target lights were placed at a

standardized distance of 130% of the outstretched right arm when holding it in

line with each target. The distance was measured between the subject’s xiphoid

process (from where they initiated pointing movements) and the tip of the right

index finger when the arm was extended towards each target. Subjects were

asked to stand straight, maintain neutral scapular retraction during the

measurement, but were permitted to rotate the trunk such that the shoulders

were facing square to the target for leftward targets. This was done in order to

measure the distance to the targets in a postural configuration similar to that

adopted during the reaching movements. During pilot tests, it was found that

the target distance of 130% could be attained comfortably by the subjects using

a combined arm and trunk movement, but did not place them at their limits of

stability at target attainment (determined by measuring centre of pressure

displacement within the base of support). Targets were all at right acromion

height as measured during quiet stance. No constraints of accuracy were given

to subjects, only that they had to point and press the light switch at their natural

speed, immediately following target illumination.

Subjects began their reaching movements with the index finger of their

right hand pressing the chest switch centered at the xiphoid process of the

sternum and their left hand hanging vertically at the side of the body. They

were asked to reach to, press and release the target switch with their right index

finger and hold this body position for 2 s, then return slowly and assume their

initial position. The total acquisition period consisted of a 3 s window. The data

66

collection time line is illustrated in Fig. 4.1B. Once the experimenter was

satisfied that the subject was standing quietly data acquisition was begun. After

a random period of between .5 s and 1 s, a target illuminated and subjects

initiated their movements. The total acquisition period of 3000 ms was

sufficient to record the postural activity preceding and accompanying the

movements. Other than movement speed (natural speed), no other instructions

regarding the strategies to be adopted were given. Subjects were not encouraged

to move as fast as possible (a reaction-time paradigm). Testing began with an

acclimatization period, during which subjects executed 26 trials in random

order, twice to each of the targets. Following the acclimatization period,

subjects were required to execute 15 trials towards each pointing direction, in a

pseudo-randomized order, which included 15 catch trials (no target light

illuminated) to reduce the possible prediction of the upcoming target light.

Thus, a total of 210 trials per subject was recorded and used in the subsequent

analysis. Generally, subjects performed blocks of approximately 50 trials inter-

spaced with 5 min rest periods to reduce fatigue until the required number of

trials at each direction was attained.

67

Figure 4. 1: Plan view of the target array and temporal sequence of data collection. A. Subjects stood on 2 force plates, 1 under each foot and were centered in a 180° light target array,

adjustable for each subject in height and distance (see Methods). Targets (light emitting diodes,

LEDs) were placed at 15o intervals from right to left sides with the position of each LED set to

exactly 130% of their outstretched arm length at shoulder height. Fy = anterioposterior force,

Fx = mediolateral force and Fz = vertical force. B. Temporal sequence of the data collection

period. An auditory tone 500 ms in length sounded to inform subjects of an impending target

illumination. A period of 1000 ms preceded the onset of the target light upon which subjects

were required to reach and point to the target. The total acquisition period was 3000 ms. A

representation of an approximate movement length (movement time, MT) is shown.

68

4.4.4 Data analysis

Kinetic, kinematic and EMG data were recorded and stored on a PC

computer for further analysis using a series of customized programs written in

MatLab (The Mathworks, Natick, MA, USA). Ground reaction forces and

moments were low-pass filtered using a digital second-order Butterworth filter,

with a 10Hz cut-off frequency. Raw EMG signals were high-pass filtered at

35Hz, de-meaned, rectified, and low-pass filtered at 100Hz (second-order

Butterworth filter). Individual trials were visually inspected for stability during

the quiet stance period (1000 ms prior to target illumination). Any trials that

showed significant variation in Fz and/or CoP during the quiet stance phase (0-

500 ms following data acquisition onset) were eliminated from further analysis,

as were trials in which subjects did not make contact with the target, or used the

target as a support. This was evaluated by checking that the centre of pressure

(CoP) did not leave the limits of the BoS determined using the kinematic

markers on each foot in relation to foot width (included in the Vicon model)

and stance distance and whether switch depression was maintained. Following

the trial selection criteria, a total of 1437 trials from the 8 subjects were retained

for further analysis from a total of 1560 trials.

Focal movement onset and termination were established by chest switch

release and target light depression, respectively. The onset of movement using

the switch was compared to the first deflection above zero of the tangential

velocity of the marker placed on the hand in the Y-axis and movement

termination was also taken as the moment when hand velocity returned to zero.

As no difference greater than 5 ms was found between the onset of motion as

determined by the chest switch and the motion capture data (sampling 1 image

approximately every 5 ms), movement time was taken as the duration between

chest switch release and target depression. Onset of focal movement acted as

time zero upon which all preparatory events were based.

To investigate the temporal and spatial organization of feedforward

preparatory postural adjustments prior to movement onset (hereon called the

69

pPA period) and associated postural adjustments produced during the end phase

of the reaching movements (the aPA period), the analysis was restricted to 250

ms before chest switch release (pPA) and 250 ms before the end of the

movement (aPA). The choice of the period lengths was based upon well

documented changes in anticipatory postural adjustments preceding voluntary

movement, or the pPA period (Belenkii et al. 1967; Bouisset and Zattara 1981;

Crenna and Frigo 1991) and visual inspection of the onset of braking forces and

moments exerted under each foot before the end of the movements for a

number of trials in each direction (aPA period). Both of these periods were

divided into five 50 ms long ‘bins’ chosen to characterize the evolution of the

preparation and the termination of reaching movements (e.g., pPA1, pPA2,

etc.). EMG amplitudes were calculated as the mean of each 50 ms bin for each

period and for each muscle. For each period (pPA and aPA) and muscle, the

highest mean response across the 13 reach directions was used to normalize

muscle activations, so that values ranged between 0 and 1. Normalized

amplitudes were then plotted as muscle tuning curves. Amplitude and direction

of the resultant horizontal force were calculated by summing changes in Fx and

Fy, according to trigonometric vector addition (Zar 1999). Resultant horizontal

plane vectors were plotted in polar coordinates. The torque about the vertical

axis at each foot (Tz) was calculated using the following equation:

Tz = Mz − Xp • Fy + Yp • Fx

where Xp and Yp are the coordinates of the centre of pressure of each

foot, Mz is the moment of the force plate around the vertical (z) axis, and Fy

and Fx are anterior-posterior and mediolateral forces, respectively.

4.4.5 Statistical analysis

The effect of pointing direction upon movement time was examined

using a repeated measures one-way analysis of variance (ANOVA). To

determine if a force constraint strategy characterized the horizontal ground

reaction forces in each of the periods under study, the angle of the average

vector produced in successive 50 ms bins was pooled from all subjects and

70

subjected to circular statistical analysis. Circular statistics refers to a class of

techniques developed for the analysis of directional or cyclic data, assuming an

arbitrary zero and where 0° and 360° represent the same direction. To

determine if pooled vector directions were uniformly distributed or constrained

(defined as a significant clustering of the active horizontal force vector,

Macpherson 1988a), data were examined for a bimodal distribution using the

‘broken- axis approach’ (Holmquist and Sandberg 1991). This approach tests

the null hypothesis that populations of vector direction are uniformly

distributed around a circle against the working hypothesis that they are not so.

Specifically, it evaluates bi-modality, but does not assume that means of

clusters of vectors are separated by 180°. If a bimodal distribution is found, it

returns the mean angles of the two modes. Typically, when analyzing circular

data sets, to evaluate the degree of dispersion mean vector length (r) is

computed. Values may vary between 0 (high dispersion) and 1 (all data

concentrated along a single preferred direction). In the broken axis approach,

rmax represents the mean vector length (r) where the modality of the data set

(k) best fits the data. The value of k specifies the type of modality and no

assumption is made that the data is uni- or bimodal. For example, if k=2, the

data would be symmetrically bimodal (i.e., 2 clusters separated by 180°).

However, when k lies between 1 and 2 the data set would be characterized by

an intermediate clustering of a bimodal distribution where clusters are separated

by 360°/k (e.g., when k=1.6, modes are separated by angle of 225°). This

approach also gives the mean direction (alpha) of any number of clusters that

can be identified. All statistical analyses were custom written and performed in

Matlab.

4.5 RESULTS

4.5.1 Kinematics of reaching movements during standing

Examples of the general kinematic strategies adopted by subjects when

pointing to targets in the array are illustrated in 3D in Figs. 4.2A-C. Three

71

principle directions of reach are shown for 1 representative subject. For all

directions, subjects began from a similar initial quiet stance position with the

right finger pressing the chest switch and the left hand by the side of the body

regardless of reaching direction. The similarities of the initial CoM positions at

movement onset can be seen in Fig. 4.2D (open colored squares). For

movements to 0°, subjects executed a rightward extension of the arm, a

clockwise rotation of the torso and a slight rightward displacement of the pelvis

(Fig. 4.2A). For movements to 90° (Fig. 4.2B) the reaching arm outstretched

approximately along the body mid-line, while the torso rotated forwards and

slightly downwards. Finally, for far leftward movements (Fig. 4.2C), the right

arm rotated leftwards, crossing the mid-line, with the torso and pelvis also

rotating towards the left. For each of the 3 movements shown, the CoM was

displaced from its initial position in the approximate direction of the target

within the base of support. Ground reaction force vectors during the pPA period

(red vectors) were oriented so that the CoM moved in the direction of each

target. At the end of the aPA period the GRF vector was oriented back towards

the CoM (blue vectors) to brake the movement of the body. Although there was

some variability (especially between 90° and 120°), trajectories and end CoM

positions differed for each reaching direction. This is represented for 5 trials for

each direction in Fig 4.2D. Despite the greater amount of body rotation required

by rightward movements, movement times did not significantly increase across

the 13 directions. Average movement times for all subjects are shown in Table

4.I.

72

Table 4. 1 Mean (±1SD) of movement times for pointing movements in all 13

directions

.

73

Figure 4. 2: 3D kinematic representations of reach to point movements to 3 principal target

directions. (A. 0°, B. 90° and C. 180°) for 1 subject (S5). Stick figures are shown as if being

viewed from the front and slightly to the subject’s left (A) and right (B and C) sides. Body

movements are shown at 2 different times: At the onset of finger movement (grey sticks) and at

the end of the movement (black sticks) when the finger touched the target. Shown are the

following segments bilaterally: foot, shank, thigh, shoulder, upper arm and lower arm. Markers at the level of the 7th cervical and 10th

thoracic vertebrae, the clavicle and sternum form a single

segment that represents the torso in the sagittal plane. The head is represented by markers

placed at 4 locations on the left and right temples and at the same level at the back of the head.

Finger trajectory is shown in red. The body centre of mass (CoM) is shown as a grey and black

circles (onset and end of the movements, respectively). The ground reaction force vector is

represented at the onset of movement in red, and at target attainment in blue. D. Trajectories of

the CoM from the onset of the focal movement (open squares) to when the finger touched the

target (open circles). Trajectories are represented for 5 trials in each direction and are colour-

coded as per the legend across the bottom of the figure.

74

4.5.2 EMG activity in relation to the forces produced: pPA period

There was a clear modulation of EMG activity and force between 0 and

180° target directions during the pPA and aPA periods. Changes in muscle

activity and force for 3 main directions of reaching are shown in Figs. 4.3A-C,

for 1 representative subject. For clarity, the two 250 ms time periods under

study have been shaded. During the pPA period, regardless of pointing

direction, the earliest change in EMG activity was typically an inhibition of

either the right or left soleus muscles. Postural adjustments for far rightward

movements (Fig. 4.3A) began with an inhibition of Solr followed by an

activation of TFLl. This was followed by a leftward and backward push at the

left leg and an unloading of Fz under the right foot. The torque about the

vertical axis (Tz) showed that the left foot exerted a counter-clockwise (CCW)

moment, while the right foot exerted a clockwise (CW) one. Centrally-oriented

movements (Fig. 4.3B) showed a clear bilateral soleus inhibition/tibialis

anterior activation and both peroneus muscles activated, which likely assisted

the initiation of the forward displacement of the body. The TFLl and RFl also

showed anticipatory bursts, as did the BFr. This may have reflected the slightly

asymmetric nature of the movement using the right arm and hand reaching to

the 90° target. There was very little or no change in Fx under either foot but

instead an increase in Fy (backwards push) and Fz (loading) under the right

foot, coupled with a slight forwards push at the left foot. As these movements

involved essentially a forward displacement, there was only a slight CW Tz at

the left foot. For movements to the far leftward 180° target (Fig. 4.3C), which

required the greatest amount of body rotation, there was significant anticipatory

activity in the BFr. Additionally, there was a burst in the Solr. The forces

showed that the right (loaded) foot Fx pushed rightwards exhibiting a CW Tz.

Interestingly, although unloaded, the left foot also showed a CW Tz.

75

Figure 4. 3 Electromyographic activity, changes in force and vertical torque (Tz) under each

foot for reaching movements to 3 principal directions (A. 0°, B. 90° and C. 180°). Traces are

shown for a period of 500 ms preceding movement onset until the end of each movement for

one typical trial in subject 5. On each plot, the full grey vertical line indicates the onset of the

light target (Light on). The dashed grey line to the left of movement onset (Movt on) indicates the onset of force and Tz changes during the pPA period. The dashed grey line to the right of

movement onset (between 500 and 750 ms) indicates the end of the arm movement (Movt end).

Forces are shown as forces exerted against the ground. TFLr and TFLl = tensor facia latae

muscles (right and left, respectively), RFr and RFl = rectus femoris, BFr and BFl = biceps

femoris, GasLr and GasLl = gastrocnemius lateralis, , Perr and Perl = peroneus longus, TAr and

TAl = tibialis anterior, Solr and Soll = soleus. Fx = mediolateral force, Fy = anterioposterior

force, Fz = vertical force and Tz = torque exerted around the vertical axis. Left =leftwards, back

= backwards, load = loading and CCW = counterclockwise. Left and right foot forces are

represented by solid and dashed traces, respectively (see legend). Shaded regions represent the

2 periods of 250 ms under study (pPA and aPA). Successive squares underneath the time axes

pictorially represent changes in Tz at each foot during the two periods.

76

4.5.3 EMG activity in relation to the forces produced: aPA period

During the aPA period, from 250 ms before to the end of the

movements, there were also distinct patterns of muscle activity across the

reaching directions, mostly in the extensors. Rightward (0°) movements (Fig.

4.3A) were characterized by activation of the GasLr, Solr, TAr and Perr. Apart

from the slight activation of TFLr early in this period, there was little activity in

the other muscles. Forces showed that the right foot was loaded, while the

horizontal forces stabilized around their original values. The loaded (right) foot

exerted a CCW Tz. Reaches to 90° (Fig. 4.3B) showed bilateral activity in a

number of extensor muscles (e.g., gastrocnemius, soleus, biceps femoris and, to

a lesser extent, peroneus). The right foot was loaded and pushed forwards (Fy),

while Fx forces steadied. The loaded (right) foot exerted a CCW Tz during this

period, but which reversed shortly after the hand reached the target (see arrow

and asterisk to the right of Movt End). Finally, reaches to 180° (far leftward,

Fig. 4.3C) showed activity in left-side extensors (GasLl and Soll) as well as

right side flexors (BFr and TAr). The loaded left foot pushed in a slightly

rightward and backward direction exerting a CCW Tz.

4.5.4 Feedforward postural adjustments show directional tuning and

are synergic

There was a clear directional tuning of many of the postural muscles

recorded during the 2 periods of study (see Fig. 4.4). In the pPA period,

patterns of EMG activity showed that the tibialis anterior and peroneus, as well

as the rectus femoris (all bilaterally) were activated principally for pointing to

mid-range targets, between approximately 45° and 120°. Other muscles, such as

the tensor facia latae, and soleus showed reciprocal patterns of activity between

reaches to the left and right. The TFLl activated between 0° and 75°, the TFLr

between 105° and 180°, while the Soll displayed inhibition during the pPA

period between 45° and 180° and sustained activity for 0° to 30° movements, as

77

did the Solr between 150° and 180°. The gastrocnemius muscles showed no

activity in the preparatory period, except for GasLl at the extreme rightward

pointing direction (0°). Other muscles, such as the BFr displayed asymmetric

patterns of activity, activating between 60° and 180° on the right side, but not

on the left.

In the aPA period reciprocal patterns of activity could also be noted for

the extensors soleus and gastrocnemius, left side muscles activating for targets

75-180° and those on the right side from 0- 90°. The BFl also activated for 75-

180° targets, while other left side muscles remained largely inactive. On the

right side, the flexor Perr contracted along with Solr as did the TFLr to a

certain extent. Interestingly, BFr showed activity across virtually all reaching

directions during this period

78

Figure 4. 4 Representative EMG traces for 14 selected muscles for S5 across the 13 directions

of pointing. Muscle activity is shown for a total duration of 500 ms, 250 ms before and after the

onset of the pointing movement. Muscle name conventions are as described in Figure 4.3. The

shaded area to the left of time zero on each muscle plot represents the 250 ms preparatory

period. Unless shown, muscles have the same scaling for the left leg (top row) as they do for the

right leg (bottom row).

79

To evaluate the spatial and inter-trial variability characteristics of the

muscular patterns described above, tuning curves were generated for each of the

14 postural muscles recorded during 5 successive 50 ms bins. For both pPA and

aPA periods, each muscle was activated for a range of pointing directions with

many showing directions of maximal recruitment. Typically, during the pPA

period (black lines and open circles) in bins 1 and 2 (pPA1 and pPA2,

respectively) all muscles were generally inactive or showed a baseline level of

activity across directions, hence, for clarity, they are not depicted in Fig. 4.5.

However, in pPA3 (-150 to -100 ms) a directional tuning began to emerge,

despite the evident inter-trial variation that existed. During pPA4 and pPA5, the

tuning curves show that the muscles worked in groups, with TAr/TAl, Perr/Perl

and RFr/RFl activating maximally for targets either side of the central one,

TFLr, BFr and Solr for leftward targets and TFLl and GasLl for rightward

movements. Clear patterns of tuning were also evident for all bins of the aPA

period (grey lines and filled circles) with the patterns emerging from bin 1

onwards (250 to 200 ms before target attainment). Again, for clarity and due to

the constancy of tuning across the 5 bins in this period, only bins 3-5 are

depicted in Fig. 4.5. The muscles Solr, GasLr, Perr and TFLr activated for

similar directions, as did GasLl, Soll and BFl. The BFr showed increasing

activity from 0-180°.

This pattern of muscle tuning was very consistent across the 8 subjects

tested. This can be seen in Fig. 4.6. Here, muscles are depicted for the final bin

of the pPA period (Fig. 4.6A) and aPA period (Fig. 4.6B) and are grouped

according to the directions for which they were recruited to a similar extent.

Figure 4.6C summarizes the recruitment of each of the muscles in the groups

identified in the 2 periods in polar coordinates. The spatial pattern of each

group represents approximately the outermost limits of the EMG tuning.

80

Figure 4. 5 Muscle tuning curves for the EMG activity of all 14 postural muscles during the

final 3 bins of the preparatory and associated periods for the representative subject S5.

Differences in tuning and recruitment of the muscle studied can be observed by comparing the

activity of the muscles over the 3 5 equivalent bins (left to right columns). Dots indicate

amplitudes from each trial measured and the solid lines the mean responses. Muscle name

conventions are as described in Figure 4.3.

81

Figure 4. 6 Muscle tuning curves in the final bin of the pPA (A) and aPA (B) periods for each

of the 8 subjects studied. Muscles have been grouped into the 3 major groups that activated for similar directions of reach. Tuning curves and individual trials are represented as in Figure 4.5.

Schema (C) summarizing the approximate range of directions of reach to which each identified

group contributed.

82

4.5.5 Spatial patterns of force differ between preparatory and

associated postural adjustments

Horizontal GRFs displayed distinct and opposing patterns for the 2

periods studied. The forces produced during the pPA period approached more

of a bi-modal distribution with low dispersion of the vector directions, whereas

those produced in the aPA period, although not unimodal, were more highly

dispersed, suggesting that to control final body position for each direction of

reaching, a unique vectors of force were produced under each foot.

These trends are illustrated in Fig. 4.7 for a representative subject. All

trials for this subject are plotted as horizontal vectors for each successive bin of

50 ms for each period. In the early phases of the pPA period (Fig. 4.7A, pPA1

and pPA2) horizontal forces were of very small amplitude and showed a high

degree of dispersion in terms of their direction. However, from the bin pPA3

onwards, which corresponded approximately to the emergence of EMG tuning

described above and shown in Fig. 4.5, forces, in particular under the loaded

foot shared similar directions (e.g., 0° to 75° for the left foot and 105° to 180°

for the right foot). Clearly constrained directions of force can be seen under the

left foot in pPA4 and pPA5, which was loaded for targets 0°-90° pushing

outwards and leftwards (see different black-blue colored vectors) and under the

right foot, loaded for 105°-180° targets, pushing outwards and rightwards (see

yellow-orange vectors), although right foot vectors were always slightly more

distributed. For clarity in Fig. 4.7A, the black arrows convey the general

direction of the exerted force in the constrained vectors when the limb was

loaded limb and grey arrows when it was unloaded. The constraint observed in

the horizontal vectors during pPA4 and pPA5 is supported by the circular

statistical analysis (see below). Interestingly, the forces produced under the foot

that was unloaded during these bins often also showed a constraint of horizontal

force under the left foot, but was slightly more dispersed under the right.

83

During the aPA period, the feet that were loaded for their respective

directions of reach during the pPA period now became unloaded and vice versa.

In bins pPA4 and pPA5 loaded vectors exerted force backwards and outwards

at the foot contra-lateral to the direction of reach (e.g., left foot, pPA5, blue

vectors). In comparison, during the aPA period, loaded vectors now pushed

outwards and forwards at the limb ipsilateral to the direction of reach (e.g.,

aPA5, orange-red vectors). Such a low dispersion of vectors seen during the

latter stages of the pPA period (especially under the left foot) was not seen

during all bins of the aPA period. Horizontal force vectors produced under the

loaded or unloaded feet in the aPA period were highly distributed, often

throughout a 90° range or greater (see Fig. 4.7D).

84

Figure 4. 7 Individual resultant horizontal ground reaction force vectors and average values of

Fz produced during the pPA period (A, B, respectively) and the aPA period (C, D) for subject 5.

Forces are shown for each consecutive bin during each period in successive rows from top to

bottom. Black and grey arrows represent the approximate direction of exerted force under the

loaded and unloaded feet, respectively. In B and D, bars above the top of the plots marked ‘L’

indicate directions of reach for which Fz under each respective foot was loaded. For reference,

the directions of reach used are indicated on the first plot (left foot) for the pPA1 period in Fig.

4.6A.

85

Circular statistical analysis (for an explanation, see Methods) supported

this trend of a force constraint strategy during the pPA but less so during the

aPA. The statistical analysis resulting from the broken axis approach is given in

Table II for both feet. During the pPA period, the analysis identified an angular

distribution that was significantly different from uniform for progressive bins.

The distribution of horizontal force was characterized by a clustering of vectors

with k values that increased during pPA4 and pPA5 to >1.5 for the left foot and

1.35 and 1.42, respectively, at the right foot. This indicated that an asymmetric

bimodal distribution existed during those periods. Additionally, rmax values

were > 0.5 (e.g. 0.66 and 0.77 for the left foot and 0.55 ad 0.6 for the right foot

during bins pPA4 and pPA5 respectively) indicating a low degree of dispersion.

Such an evolution of significant bimodal clustering was not seen during the

aPA period. Rather, statistical analysis, revealed that the distribution of force

vectors was significantly different from uniform with rmax values ≤0.5,

indicating higher dispersion than in the pPA period around a single mode.

86

Figure 4. 8 Average direction and magnitude of horizontal ground reaction force change during

each bin of the pPA and aPA periods under both feet.

87

Table 4. 2 Results of the broken-axis approach (Holmquist & Sandberg, 1991)

n = 1437, k=degree of modality (1 = unimodal, 2 = bimodal, two clusters separated by 180°),

rmax = mean vector length for which k best fits data, alpha 1 and alpha 2, = mean angle in

degrees of cluster 1 and 2 respectively, identified using the broken-axis algorithm and p =

probability that the distribution of force vector angle is significantly different from uniform. NS

= not significantly different from a uniform distribution.

88

The bimodal distribution of force vector direction highlighted by the

circular statistics in the pPA period is also reflected when average vector

direction is plotted against each reach direction (Fig. 4.8A). It can clearly be

seen that there was a constraint of force vector direction during pPA4 and

pPA5, especially under the left foot. This constraint of force vector direction

seen in the pPA period was also accompanied by a modulation of force

magnitude with targets at each extremity having the largest GRF (Fig. 4.8B,

pPA4 and pPA5), whereas the aPA period displayed a linearity of GRF vector

and reaching direction from bin 2 onwards (Fig. 4.8C) and a fairly constant

magnitude of GRF vector across all directions (see Fig. 4.8D, all bins).

4.6 DISCUSSION

We investigated the organization of feedforward postural adjustments

produced when standing humans reached with their preferred arm to multiple

targets placed in a semi-circle throughout 180° with centered respect to their

midline. Our objective was to identify spatial tuning of postural muscle activity

and a constraint of horizontal GRF, similar to that shown for feedback- based

postural responses to unexpected perturbations of balance in humans (Henry et

al. 2001) and animals (Macpherson 1988a). The results supported our first

hypothesis; that bilateral EMG activity recorded was directionally-tuned and

served the current goal of the task (initiating or terminating the movements).

However, our second hypothesis; that a generalized force constraint strategy

existed for feedforward postural adjustments, was not completely supported.

4.6.1 The roles of preparatory and associated postural adjustments

for reaching during stance

The postural adjustments studied in the two periods likely performed

functionally different roles. In the first period (pPA), significant EMG activity

emerged approximately between -150 ms to movement onset. The resultant

89

horizontal forces produced under the loaded foot during this period were

consistently in the opposite direction to the desired movement direction, the

vectors being constrained for a number of reaching directions. Moreover, the

moments (Tz) produced indicated that the loaded foot consistently exerted a

torque in the opposite direction to the direction of reach (Fig. 4.3), thus creating

a reaction torque that drove the body towards the target. These results would

support the notion that the pPA created the necessary conditions for CoM

displacement in the direction of the reach, within the base of support (Stapley et

al. 1998; Stapley et al. 1999). This is also supported by the CoM trajectories in

Fig. 4.2D. The pattern of force and torques exerted at the feet reversed during

the aPA period, such that the reaction forces opposed the body displacement as

the hand reached the target. Perhaps the clearest example of this could be seen

under both feet for 180° reaches (Fig. 4.3C), which required the greatest

amount of body rotation. The CCW direction of Tz would have had the effect

of slowing and stabilizing CoM position within the support base as the hand

neared the target. Such a role of associated postural adjustments has been

shown during various arm reaching or lifting tasks (Commissaris and Toussaint

1997; Commissaris et al. 2001; Cordo and Nashner 1982; Krishnamoorthy and

Latash 2005; Schepens and Drew 2003).(Macpherson 1991)

4.6.2 Tuned, synergic muscle activity characterizes feedforward

postural adjustments

The activity of the 14 muscles recorded in the pPA and aPA periods

showed that each muscle was recruited for a range of directions often with a

direction of maximal activation. Muscles were also activated in groups. This

grouping across the two periods would support a synergic organization of

feedforward muscular activity. A synergy has been defined as a group of

muscles constrained to act in a concerted manner (Macpherson 1991;

Sherrington 1961) or ‘activated in synchrony with fixed relative gains and

muscle activation patterns with consistent spatial characteristics’ (Torres-

90

Oviedo and Ting 2007). Directionally tuned feedforward muscle activity has

previously been identified during bilateral arm movements performed when

standing (Aruin and Latash 1995a). When the two arms were moved either

forwards or backwards from a central position, groups of muscles were

activated in functional groups on the dorsal or ventral side of the body to

maintain equilibrium. Direction-specific patterns of feedforward muscle activity

have also been identified in dorsal muscles in anticipation of perturbations to

equilibrium (Latash et al. 1995; Santos and Aruin 2008). The muscle activity

identified in the present study also served a range of directions and was

maximally tuned to a specific reaching direction in both periods studied. We

suggest therefore that the patterns of muscular activation seen in both periods

may belong to the same limited number of robust functional muscle synergies,

despite the inter-trial variations seen in each subject reflecting different levels

of synergy activation. However, APRs are primarily triggered by afferent

feedback from the moving surface, whereas in the present study the postural

adjustments were produced in anticipation of the upcoming movements.

Therefore, similarities in the synergic organization of feedback and feedforward

postural adjustments may support the idea that the CNS adopts functional

synergies for both modes of control to produce different motor behaviors (Ting

2007).

4.6.3 Clearly constrained force patterns are seen during preparatory

but not during associated feed- forward postural adjustments

A force constraint strategy has been defined as a bimodal clustering of

active force vectors where the forces are constrained to act along an

approximately diagonal axis directed roughly toward or away from the CoM at

a single limb (Macpherson 1988a; 1994). Until now, this strategy has been

identified in the horizontal forces comprising the feedback-based postural

adjustments produced when the surface upon which animals or humans were

standing was unexpectedly displaced (Henry et al. 2001; Macpherson 1988a;

91

1994). Our results have shown statistically, that a force constraint strategy

existed for feedforward postural adjustments accompanying a pointing task,

specifically during the latter stages of the pPA period (pPA4 and pPA5) during

which the objective was to displace the CoM in the direction of the target.

During these bins the angular distribution was significantly different from

uniform and rmax values (quantifying the degree of dispersion, see Methods)

supported a low dispersion of vector direction. Our rmax values were

comparable to, if not greater than, those produced by cats perturbed in the

horizontal plane standing at natural stance width (Macpherson 1994). Average

values of k during these same periods (pPA4 and pPA5) were between >1 and

<2, indicating that an asymmetric bimodal distribution existed (less than 180°

separation existed between clusters). We conclude, therefore, that a force

constraint strategy existed for pPA’s. Such an observed pPA period force

constraint strategy may represent a high-level task variable whereby the CNS

groups together the dynamic forces required to execute a number of directions

of reaching in order to simplify the complex control of multijoint movement.

It remains, however, that such a low dispersion of force vector was not

observed in our study for the forces exerted during the aPA period. Values of

rmax were <0.5 (more highly dispersed) and only one single alpha value was

identified. This more dispersed force pattern observed during the aPA period

when compared to the pPA period may be explained by the current nature of

the task. Task instructions were to maintain final finger position at target

position and not to return to the initial position. Thus, the constraints of

displacing the CoM from one position to another, as in the pPA period, did not

apply. Rather, subjects had to maintain their body position and stability at target

attainment. Force constraint strategies have been documented to exist mostly

when humans or animals are required to actively displace the CoM back

towards its initial position following a perturbation (Henry et al. 1998a; 2001;

Macpherson 1988a; Ting and Macpherson 2004). In this instance, the force

constraint strategy would simplify neural control mechanisms to coordinate

force direction and amplitude during these ‘active’ responses (McKay et al.

92

2007). During the aPA period of this study, the task was simply to decelerate

the body to target position and remain in that position. It did not require a

similar ‘active strategy’ of CoM (and body) displacement, but rather a

maintenance of posture. Thus, we suggest that the force constraint strategy

characterizes feedforward postural adjustments when the goal is to displace the

CoM within the base of support (likened to the active adjustment following a

perturbation) but not when task requirements are to precisely control an end

posture and finger position during voluntary reach to point movements.

Finally, it may be asked why the force constraint strategy observed

during the pPA period was weaker at the right foot than at the left? The right

foot, ipsilateral to the reaching arm, showed a less constrained pattern of

horizontal force during pPA4 and pPA5. It is possible that the left foot

(contralateral to the reaching arm) was primarily responsible for producing the

turning moment to targets 0-90°. Indeed, Fig. 4.3A shows that when loaded, the

left foot produced a CCW Tz, while the right foot opposed that torque. The

reverse was not true, however for movements to 180°. Here, the left foot

(although unloaded) exerted a Tz that was in the same direction as the right to

drive the body around to the left. Moreover, at this direction during the aPA

period, the right foot consistently opposed the direction of reach, likely to assist

the braking of the movement. Thus, we tentatively suggest that the leg

contralateral to the arm assists in turning to its contralateral side, whereas the

leg ipsilateral to the reaching arm is perhaps more coordinated with the upper

limb and establishes a base for the limb’s trajectory during the reaching task.

Another possibility is that as all subjects were right-handed, their right leg (and

foot) was also their dominant one. If so, it may be that the dominant limb is

better adapted to controlling the limb’s trajectory by producing more dispersed

vectors, while the left leg ensures a postural transition at the onset of reaching

directions for which it is loaded. A differentiation of trajectory control versus

steady-state limb posture has been identified for dominant and non-dominant

arms (Duff and Sainburg 2007). Our further experiments will vary the reaching

arm used in an attempt to confirm these hypotheses.

93

4.6.4 Implications for the neural control of balance: shared control of

feedforward and feedback postural adjustments

This study has shown that EMG tuning characteristics in the pPA period

are organized in a similar manner to compensatory postural responses produced

via feedback. This raises an important question: To what extent do similar

neural pathways control posture during voluntary movements and following

unexpected perturbations? It is well known that central commands for

feedforward and feedback postural adjustments originate at supraspinal levels.

Commands for APAs arise in different regions of the cortex (Gurfinkel and

Elner 1988; Massion et al. 1989; Wiesendanger et al. 1987). In addition, spinal

cats have difficulty maintaining equilibrium when the support surface is

unexpectedly displaced horizontally (Macpherson and Fung 1999) and do not

exhibit the complex patterns of evoked EMG characteristic of APRs in the

intact animal (Macpherson et al. 1997). Lesion studies have shown that damage

to the brainstem reticulospinal system impairs balance (Gorska et al. 1990;

Gorska et al. 1995; Lawrence and Kuypers 1968). In addition, the injection of

cholinergic agonists, which excite reticular neurons or noradrenergic agents that

block inhibitory neurons to the reticular formation affect feedforward

programming of postural responses accompanying reaching (Luccarini et al.

1990). More recently, it has been shown that signals in the brainstem

pontomedullary reticular formation (PMRF) contribute to feedforward postural

adjustments during reaching with neurons in the PMRF encoding either pPAs

or aPAs, or both (Schepens and Drew 2006; 2004). It has also been shown that

microstimulation of peripheral afferents, known to be essential for the early

triggering of APRs (Stapley et al. 2002), also activate particular subsets of

reticulospinal neurons (Drew et al. 1996). Moreover, the activity of RSNs in the

same anatomical area of the PMRF as those related to feedforward postural

adjustments (Schepens and Drew 2006; 2004), also contribute to the

compensatory postural responses recorded in the limbs following unexpected

94

perturbations (Stapley and Drew 2009). Thus, evidence would suggest that an

integration of feedforward and feedback modes of control may occur in the

brainstem.

What would be the end result of commands sent from the brainstem to

the muscles producing both types of postural adjustments? Recent evidence has

shown that, rather than a unique muscle synergy existing for a particular

direction of postural perturbation, APRs are organized into a few muscle

synergies which represent the general neural strategy that accounts for the

spatio-temporal components of the response (Torres-Oviedo and Ting 2007).

These muscles synergies, or primitives, may be used to control task level

variables such as CoM motion or displacement (Ting and Macpherson 2004) or

CoP displacement (Krishnamoorthy et al. 2003). We propose that the similarity

in organization of preparatory (feedforward) postural adjustments and

compensatory (feedback) postural adjustments may take place in the PMRF.

This structure would then organize the functional muscle synergies required in

either mode of control and the outward expression of those synergies, (the

active force vectors) required to produce whole body movements or active

corrections to balance. The results of this paper do not, however, enable us to

conclude that feedforward and feedback based postural adjustments share

common ‘motor primitives’. Further study is needed therefore, to characterize

the muscle synergy organization during feedforward and feedback postural

tasks in the same muscles of the same subject.

4.7 ACKNOWLEDGEMENTS

This study was supported by a Canadian Foundation for Innovation

(CFI), New opportunities fund grant (#10613) and a Natural Sciences and

Engineering Research Council (NSERC), Individual discovery grant to Paul J.

Stapley. The authors acknowledge the expert technical assistance of JJ Loh. We

also thank Drs. Jane Macpherson and Trevor Drew for helpful discussions

about earlier versions of this work.

95

Chapter 5

Muscle synergy characterisation of

feedforward postural adjustments during

reaching in standing humans

5.1 PREFACE

Chapter 4 investigated the spatial organization of feedforward postural

adjustments during reaching to targets in multiple directions while standing. It

was found that while muscle activity was directionally-tuned to the direction of

reach, forces exerted at the ground were constrained in direction, with an

anisotropic relationship observed between GRF direction and reach direction.

These characteristics bear striking resemblance to the spatial organization of

feedback-based postural responses to unexpected support surface translations in

cats (Macpherson 1988a; b) and humans (Henry et al. 2001; 1998b). In

feedback postural control, the variability in muscle activity has been explained

with muscle synergies. Therefore, we postulated that a limited number of

muscle synergies may underlie the patterns of muscle activity during

feedforward postural adjustments. Thus, to examine this question further,

Chapter 5 investigates whether muscle synergies underlie the feedforward

preparatory postural adjustments (pPA) occurring prior to the onset of multi-

directional reaching movements performed in the standing position.

This manuscript has been prepared for submission to the Journal of

Neurophysiology.

5.2 ABSTRACT

This study investigated whether a limited number of robust muscle

synergies underlie the observed patterns of feedforward postural muscle activity

for reaching movements during stance. We hypothesized that the recruitment of

96

a few muscle synergies could account for the spatiotemporal and inter-trial

variability of preparatory postural adjustments for multidirectional reaching

movements in the standing position. Nine subjects were asked to point to 1 of

13 light targets spaced at 15° intervals from their right to left sides (centered at

midline) while standing on two triaxial force plates. The EMG activity of 14

postural muscles were recorded and quantified in 5 time bins of 50 ms

preceding the onset of the focal movement (pPA period, presumed to be

entirely feedforward in nature) and subjected to a non-negative matrix

factorization (NNMF) analysis (Lee and Seung 1999; Torres-Oviedo and Ting

2007). For the six subjects analyzed, the temporal and spatial patterns of pPA

were well represented by the recruitment of 8 or less muscle synergies, as well

as inter-trial variability of muscle activity. Muscle synergy recruitment was

directionally tuned with respect to reach direction. While these results show

similar features in number of synergies and spatial organization to those

observed for automatic postural responses, robustness of muscle synergies for

feedforward posture control remains to be determined by evaluating whether

the same synergies are preserved for the feedforward postural adjustments

accompanying the reaching movement. Overall, these findings provide a first

step in examining whether a set of robust muscle synergies can be identified

across both feedforward and feedback modes of postural control.

5.3 INTRODUCTION

Voluntary movements executed in the standing position, such as raising

an arm or reaching out to an object, are accompanied by postural adjustments

that occur before and during the execution of the action to stabilize the body

against the perturbing effect of the goal-directed movement (Horak and

Macpherson 1996; Massion 1992). Much of what is known about the

organization of these feedforward or predictive postural adjustments, which

occur in advance or concurrent with voluntary movements, comes from the

study of single arm or leg movements (Massion 1992; Mouchnino et al. 1992).

This previous work has focused largely on the temporal aspects of postural

97

muscle activity in relation to the onset of muscle activity initiating the

voluntary component of the tasks, which were executed primarily along a single

plane. However, everyday movements such as reaching to objects, frequently

involve lateral movements of the arms to the side or across the body midline.

The postural adjustments required to initiate and stabilize posture during the

execution of reaching movements in multiple directions have been studied by a

number of authors. Results have shown that complex patterns of muscle activity

in the legs and trunk that make up functional groups which are spatially ‘tuned’

to the direction of the reaching movement (Aruin and Latash 1995a; Leonard et

al. 2009; Santos and Aruin 2008).

Spatial tuning of postural responses has been demonstrated for

feedback-based postural responses following support surface perturbations

(Carpenter et al. 1999; Henry et al. 2001; 1998b; Macpherson 1988b; Ting and

Macpherson 2005; Torres-Oviedo and Ting 2007; 2010). These automatic

postural responses (APR) represent coordinated patterns of muscle activity

organized to accelerate the center of mass (CoM) to a position of stability with

respect to the base of support (BoS) (Chiel et al. 2009; Lockhart and Ting 2007)

and are shaped by afferent feedback that signal the characteristics of the

postural disturbance (Stapley et al. 2002; Ting and Macpherson 2004). Detailed

analyses of APR have revealed that only a small number of muscle synergies

are necessary to explain the temporal and spatial variability of the complex

patterns of observed postural muscle activity (Ting and Macpherson 2005;

Torres-Oviedo et al. 2006; Torres-Oviedo and Ting 2007). It has been

postulated that the recruitment of muscle synergies simplifies the task of

controlling balance by enabling the CNS to send a single command to recruit a

group of muscles, rather than controlling each muscle individually (Cheung et

al. 2005; Ting 2007; Tresch et al. 1999).

A modular organization of the musculoskeletal system has been

proposed as a mechanism by which the CNS simplifies the control of

movement in both postural and goal-directed motor behavior (Ting 2007; Ting

and Mckay 2007). In this framework, muscle synergies, or M-modes

98

(Krishnamoorthy et al. 2003; Krishnamoorthy et al. 2004), represent groups of

muscles that are constrained to act together (Macpherson 1991; Ting and

Macpherson 2005; Torres-Oviedo et al. 2006; Torres-Oviedo and Ting 2007)

and recruited by the CNS to produce predictable biomechanical functions

(Chiel et al. 2009; Chvatal et al. 2011; Ting and Mckay 2007). The concept of

muscle synergies has been applied in several species to explain the muscle

coordination patterns in a variety of motor behaviors, including balance control

(Chvatal et al. 2011; Krishnamoorthy et al. 2003; Ting and Macpherson 2005;

Torres-Oviedo and Ting 2010), locomotion (Drew et al. 2008), and seated

reaching (d'Avella et al. 2011; Muceli et al. 2010). Although these tasks are, to

some extent, characterized by feedforward muscle activity, an identification of

the synergies produced in the postural muscles during reaching in the standing

position is lacking. Such a characterization of the existence and nature of the

muscle synergies in a postural behaviour that is assumed to be entirely

feedforward in nature would provide us with a means of deducing whether

feedforward and feedback postural behaviours share similar organizational

principles despite their different modes of control.

Therefore, the aim of the present study was to record the postural

adjustments preceding reaching movements of the preferred arm in multiple

directions, so that their spatial organization can be examined, and to

characterize the postural muscle activity using a technique that has been used to

derive synergic relationships between muscles. This experimental set up,

described in detail in Leonard et al. (2009) was chosen as it is known to evoke

spatially-tuned postural muscle activity in the lower limbs. The paradigm was

used therefore to explore whether a muscle synergy control structure can

explain the variability of postural muscle activity when subjects performed

reach movements during standing in multiple directions. Based on our previous

study characterizing feedforward postural adjustments during multidirectional

reaching during stance (Leonard et al. 2009), we hypothesized that postural

muscle activity preceding the onset of goal-directed reaching movements could

be explained by a few muscle synergies.

99

5.4 METHODS

5.4.1 Subjects

Nine subjects (8 male, 1 female), free from any known neurological,

visual or orthopedic disorders were recruited from the McGill University

student population to participate in the present study. All subjects (23.2 ± 5.2

(SD) years, a mean height of 1.69 ± 0.05 m, and a mean weight of 69.2 ± 4.3

kg) were right-hand dominant. They gave their informed consent to participate

and experiments were conducted with the approval of the McGill University

research ethics board. Data from 6 of the 9 subjects is presented here.

5.4.2 Experimental apparatus and set up

The experimental set-up used for this experiment was used previously

and is explained in detail in Chapter 4 (Leonard et al. 2009). In brief, subjects

were centered in a semi-circular target light array containing 13 targets, and

performed multidirectional reaching movements with their right (preferred) arm

while standing. Following a variable delay, one of the 13 target lights,

presented in random order, would illuminate. Subjects were required to point

and depress a target switch at their natural movement speed. Upon contact with

the target, subjects released the switch immediately and stood with their arm

outstretched at the target for a duration of 2 seconds (see Fig. 4.1B). Subjects

were presented with 15 trials to each target and 15 trials where no target

illuminated (‘catch’ trials), presented in a pseudorandom order. Trials where no

target illuminated or ‘catch’ trials were included during data collection to

minimize the anticipation of the illumination of a target light, but were

excluded from further analysis. In total, 195 trials (15 trials x 13 directions) for

each subject were retained for processing and analysis.

Postural adjustments preceding the onset of the reaching movements

were quantified in terms of kinematic, kinetic (force) and EMG activity. The

data analysis presented in this study was restricted to the activity of the

following 14 postural muscles, recorded bilaterally: tibialis anterior (TAl and

100

TAr, respectively), soleus (Soll, Solr), lateral gastrocnemius (GasLl, GasLr),

peroneus longus (Perl, Perr), biceps femoris (BFl, BFr), rectus femoris (RFl,

RFr), and tensor fascia latae (TFLl, TFLr). In addition, anterior and posterior

deltoid muscle activity was recorded at the right shoulder. However, this EMG

activity was excluded from the non-negative matrix factorization (NNMF)

analysis (described in detail below). Raw EMG data were high-pass filtered at

35 Hz, de-meaned, rectified, and low-pass filtered at 40 Hz using custom-

written programs in Matlab (Mathworks, Natick, MA).

5.4.3 Data processing and analysis

The mean activity of each postural muscle was calculated during five 50

ms time bins preceding the onset of the reaching movement, as determined

from the release of a chest switch attached to the subject’s xiphoid process (for

details of chest switch, see Chapter 4). The EMG amplitude values from all

muscles were then assembled into a single, large data matrix, where each row

represented a vector of data from a single muscle for all conditions (time bin,

reach direction and trial). Each vector was organized as follows: 5 time bins x

13 directions x 15 trials for a total of 975 data points, or conditions, for each

muscle. For display purposes, each muscle was then normalized to its maximal

activity across all time bins, pointing directions, and trials so that values fell

between 0 and 1. Normalized muscle activity as a function of reaching direction

was plotted to create ‘EMG tuning curves’ (Fig. 5.2). To ensure that each

muscle was uniformly represented across all conditions and that variations in

muscle activity were weighted equally by the NNMF algorithm, each muscle

vector was normalized to have unit variance. This means that the sum of the

squared values for each row, or muscle, was equal to zero (Ting and Chvatal

2010).

5.4.3.1 The non-negative matrix factorization technique

Non-negative matrix factorization (NNMF) has been identified as a

useful computational tool for examining the coordination between a large

101

number of muscles and biomechanical variables to test hypotheses about how

the CNS might reduce the dimensionality of control (Ting and Chvatal 2010;

Tresch et al. 2006; Tresch and Jarc 2009). It is a linear decomposition technique

that assumes that a data set constitutes a linear combination of a smaller number

of elements (Lee and Seung 1999). While this principle is not new to

computational models of sensorimotor control, NNMF differs from other linear

decomposition techniques, such as PCA or ICA, in that it constrains the

identified elements to fall within a non-negative space (Lee and Seung 1999;

Ting and Chvatal 2010).

When analyzing muscle activation patterns with NNMF, it is assumed

that the observed pattern of muscle activity is the linear sum of a f(Cheung et

al. 2005)ew (Nsyn) muscle synergies (W), each activated by a synergy

recruitment coefficient (c), (see Fig 1A, Ting and Macpherson 2005; (Torres-

Oviedo et al. 2006; Torres-Oviedo and Ting 2007; 2010; Tresch et al. 2006;

Tresch et al. 1999). Using this formulation, the predicted muscle activity (M*)

can be expressed mathematically as:

M*(t) = ∑ci(t)Wi + error

ci ≥ 0 Wi ≥ 0

error =∑∑(Eij - E*

ij)2

where Wi represents the spatial pattern of muscle activity defined by

synergy i, and ci defines the recruitment of synergy i.

A spatially-fixed muscle synergy (Wi) defines a group of muscles that

are co-activated with fixed spatial scaling. Each element, or muscle, of the

vector Wi, has a fixed value between 0 and 1 for all trials, where 1 represents

the maximal level of activity. The synergy, Wi, is recruited across conditions by

a scalar, non-negative value, ci. This recruitment coefficient, ci, is hypothesized

to reflect the neural signal that activates a group of muscles with fixed relative

amplitude ratios defined by Wi. The tuning of Ci defines how the activation of

synergy Wi is modulated as a function of reaching directions and time (Chvatal

et al. 2011; Safavynia and Ting 2011; Torres-Oviedo and Ting 2007). In

summary, the spatial organization, or composition of each synergy, Wi, remains

102

fixed while the recruitment coefficient, ci, may vary in time (Safavynia and

Ting 2011). It is hypothesized that ci represents the neural command that

defines how the muscle synergy is modulated with respect to time and reaching

direction (Ting 2007; Ting and Mckay 2007).

5.4.3.2 Assumptions of non-negative matrix factorization

The assumptions of NNMF are reviewed in detail elsewhere (Ting and

Chvatal 2010). Briefly, a fundamental assumption of NNMF is one of non-

negativity; that is, the weights ci and the components Wi are constrained to be

non-negative (Lee and Seung 1999; Ting and Chvatal 2010). NNMF is a parts-

based decomposition technique; consequently only additive features of the data

set are represented. In the context of the analysis of muscle activity, each part

identified must resemble features of the overall observed muscle activity (Lee

and Seung 1999; Ting and Chvatal 2010; Tresch et al. 2006). As a result of the

non-negative assumption, it is not possible to identify correlated inhibition

patterns between individual muscles (Tresch et al. 2006). However, decreases

in the level of activation of a synergy are interpreted as the inhibition of that

synergy in relation to the other synergies (Torres-Oviedo and Ting 2007). Thus,

relative inhibitory patterns can be examined. When applying NNMF to analysis

of muscle activity it is assumed that (1) any given muscle may be recruited by

more than one synergy (Torres-Oviedo and Ting 2007; 2010) and (2) relative

levels of muscle activation are fixed within a synergy (i.e. spatially-fixed

synergies, (Safavynia and Ting 2011; Ting and Macpherson 2005).

5.4.3.3 Extraction of muscle synergies for feedforward postural adjustments

Using NNMF, spatially fixed muscle synergies (W) and their activation

coefficients (c) were extracted from a data matrix containing the muscle activity

recordings of 14 postural muscles for the pPA period only. The data matrix was

constructed with dimensions m x s, where m is the number of muscles and s the

number of samples (number of time bin x reaching condition x number of

103

trials). For each subject, the components of the data set, W and c, were found

by iterating a number of random elements via a search algorithm (nnmf

function in MATLAB) until a sufficient proportion of the variability in the data

set (at least 75%) was accounted for (Chvatal et al. 2011; Ting and Macpherson

2005). Specifically, the NNMF algorithm randomly selects non-negative

matrices W and c and modifies their composition to minimize the sum of

squared errors between the actual data (M) and reconstructed data (M* = W x

c). This technique has been used successfully to extract muscle synergies for

feedback postural adjustments in cats (Ting and Macpherson 2005; Torres-

Oviedo et al. 2006) and humans (Chvatal et al. 2011; Safavynia and Ting 2011;

Torres-Oviedo and Ting 2007; 2010), as well as in frog scratching, swimming,

and jumping (Cheung et al. 2005; Tresch et al. 1999).

5.4.3.4 Determining the appropriate number of muscle synergies (Nsyn)

The appropriate number of synergies (Nsyn) was established for each

subject by iterating the value of Nsyn from 1 to 14 and determining the least

number of synergies that could adequately reconstruct the muscle activity

observed in each time period and each trial. Goodness of fit between the actual

EMG and reconstructed EMG was quantified using a measure of the variance

accounted for (VAF) in each muscle vector. VAF was defined as 100 x

uncentered Pearson’s correlation coefficient (Torres-Oviedo et al. 2006; Zar

1999).

Several criteria for VAF were used for determining the correct number

of synergies necessary to explain the trial-to-trial and inter-trial variability in

postural muscle activity observed across all time bins and reach directions for

feedforward postural responses. First, a minimum Nsyn was selected when a

global criteria of >90% VAF for the entire data set (overall VAF) was met.

Second, additional local criteria were imposed to ensure that all features of the

data set were reproduced by the extracted synergies. These local criteria were

quantified when only a portion of the entire data set (60% of all pPA trials)

were considered. Specifically, muscle VAF and condition VAF were

104

quantified. Muscle VAF is a measure of how well the extracted synergies

explained the activity of an individual muscles across all time bins, reach

directions and trials, whereas condition VAF considered the activity of the

entire set of recorded EMG for all trials in a single reach direction and time bin

(Chvatal et al. 2011). These local criteria were met when the Nsyn was able to

account for at least 75% of the muscle VAF and condition VAF. Thus, Nsyn

was increased until all criteria were met as a minimum, and further increased if

additional muscle synergies drastically improved local fit criteria. However, if

an additional muscle synergy contributed evenly to the VAF across muscles and

reaching directions, it was assumed to represent noise in the data set and was

therefore not included (Chvatal et al. 2011; Ting and Chvatal 2010; Torres-

Oviedo and Ting 2007).

5.4.3.5 Muscle synergy analysis

The robustness of the extracted muscle synergies was evaluated by

examining how well the muscle synergies extracted from a portion of the data

set (control data, CTRL trials; 60% of all pPA trials) was able to predict the

measured EMG in the remaining trials (remaining 40% of all pPA trials).

Specifically, goodness of fit between predicted and observed EMG was

quantified with centered (r2) and uncentered (VAF) Pearson correlation

coefficients (Chvatal et al. 2011; Safavynia and Ting 2011; Torres-Oviedo and

Ting 2010). These similarity metrics are both defined as the (sum of squared

errors)/(total sum of squares) (Torres-Oviedo et al. 2006; Zar 1999). However,

r2 is calculated with respect to the mean whereas VAF is determined with

respect to zero. As such, r2 compares the shape of two curves without

considering the amplitudes of the curves, whereas VAF considers both shape

and intercept between two data sets. VAF is therefore considered more

stringent, as both shape and intercept must be very similar for high VAF to be

achieved (Safavynia and Ting 2011; Torres-Oviedo et al. 2006).

105

5.5 RESULTS

The results showed that directionally-tuned pPAs that preceded reaching

movements in multiple directions executed during stance could be explained by

a modular organization of muscle activity. Specifically, across all subjects, the

spatial, temporal and inter-trial variability of the postural muscle activity could

be reproduced by between 5 to 8 muscle synergies. Moreover, consistent

muscle synergies extracted from CTRL trials were able to accurately reproduce

the muscle tuning curves as a result of modulation of the synergy recruitment

coefficient (C).

5.5.1 Feedforward postural muscle activity is directionally tuned, but

shows variability between trials

Reaching movements performed in the standing position to targets

located beyond reach and in multiple directions in the horizontal plane were

preceded by postural muscle activity in the supporting limbs. This is shown for

three principal directions of reach (0º, 90º and 180º) in Figure 5.1. Reaching

movements were preceded by direction-specific modulation of EMG in the

supporting limbs occurring as early as 150 ms prior to the onset of the

movement. For rightward targets (e.g., 0º, see Fig. 5.1A), pPAs were

characterized by inhibitory activity in the right limb (Solr and Perr) and

activation of the extensors in the left limb (TFLl and GasLl). Centrally oriented

reaching movements (90º, see Fig 5.1B) showed pPAs to have initial inhibition

of the ankle extensors (principally the soleus muscles) followed by activation of

the flexors (bilateral TA and TFL). Finally, targets located left of the midline

(180º, see Fig 5.1C) involved activation of extensor activity under the right foot

(Solr, BFr and TFLr) during the pPA period.

106

Figure 5. 1 Representative traces of muscle activity in seven muscles recorded bilaterally

during a 500 ms period preceding movement onset (Movt On) to movement end. The pPA

period is indicated by the shaded vertical grey area to the left of movement onset.

The sensitivity of muscle activity with respect to reaching direction and

period of the pPA was quantified by examining the mean activity over a 50 ms

period for five time bins of 50ms each. Figure 5.2 shows tuning curves (mean

trace of 15 trials across directions and the variability across each of the trials)

for one subject (S011). Clear directional tuning of muscle activity occurred in a

number of muscles and became evident 150-100ms before movement onset. For

active muscles, maximal activity occurred for a small range of reaching

directions. While similarities in the shape of the tuning curves were observed

between some muscles, such as TAl and Perl, significant variability in the

timing and amplitude of muscle activity was observed across trials and reaching

directions as shown by the large standard deviations observed within a reaching

direction.

107

Figure 5. 2 Muscle tuning curves for all 14 muscles and all 5 time bins during the 250 ms period preceding the onset of reaching movements in a representative subject (S011).

108

5.5.2 Composition and tuning of muscle synergies

For all subjects, a small number of muscle synergies were sufficient to

reproduce the spatial and temporal patterns of preparatory postural muscle

activity required for initiating reaching movements in multiple directions. The

minimum number of muscle synergies able to account for at least 90% of the

variability of the entire data set (Fig. 5.3A) and at least 75% of the variability in

each muscle (Fig. 5.3B, C) across reach conditions (Fig. 5.3D) and period of

the pPA was chosen (Fig. 5.3E). A range of 5 to 8 muscle synergies/subject was

found to sufficiently reconstruct the EMG patterns across all trials and reaching

conditions, and accounted for 93.33 ± 1.4 of the total VAF.

109

Figure 5. 3 Variability accounted for (VAF) for different number of muscle synergies for the

entire data set for a representative subject (S010). To choose an appropriate number of muscle

synergies (Nsyn), the following criteria were met: (A) Overall VAF attained a threshold of >

90%; (B) VAF by muscle as a function of muscle synergy number. (C) VAF by synergy

number as a function of muscle was plotted to confirm that chosen Nsyn met the criteria of

75%. (D) VAF by direction and bin shows that directions were well characterized.

110

Each muscle synergy (Wi) specified the relative activation levels for the

muscles hypothesized to be activated together by a neural command (Fig. 5.4).

Figure 5.4 shows the composition of the muscle synergies and the temporal

modulation of the synergy activation coefficients, which were directionally-

tuned to the reaching direction. Several muscles were activated by more than

one synergy, although their relative level of activation differed. For example,

Perll was strongly activated by W4 for rightward targets (Fig. 5.4, yellow

synergy) where it likely serve to produce eversion of the left foot and assist in

pushing the body towards the right. However, Perll was also activated for W2

(Fig. 5.4, red synergy) for forward directed targets. Generally, muscle synergies

were composed of muscles from both the left and right limb (W1, Fig. 5.4).

However, groupings were typically more heavily weighted to a single limb

according to muscle function, rather than consisting of an equal distribution of

muscles of the left and right limbs (ie, equal contributions from TAr and TAl

are not observed). For example, for subject S010 represented in Figure 5.4, 6

muscle synergies were selected. During pPA1 and pPA2, the W1 (Fig 5.4, blue

synergy) consisting of extensors Solr, Soll and GasLr (see left side histograms)

were activated, likely providing antigravity function for quiet stance.

Interestingly, the contribution of this W1 (blue) reduced as the pPA period

progressed and quiet stance changed to a preparation for movement onset.

At the bin pPA4, the most obvious changes in recruitment (tuning

curves) illustrate how the synergies were modulated according to reach

direction. The W2 synergy (red histograms, left side Fig. 5.4) subserved

forward reaching targets mainly, being composed of bilateral TA, Per and some

RF. Leftward targets (105º - 180º) involved activation of W3 mainly (pink

synergy; Fig. 5.4) which consisted of RFr, TFLr and Soll mainly. However,

additional contribution from the extensor synergy (W1) was also seen for these

targets. In contrast, targets located to the right of the midline (0º - 75º) recruited

Perl, TFLl, Gasl and Soll (yellow synergy, Fig. 5.4). Initiation of reaches

towards a smaller range of rightward targets (45º - 75º) also recruited some

111

extensor activity in the left (RFl, Soll) and right (BFr and Gasr) legs (magenta

synergy; Fig. 5.4). Finally, a broadly tuned synergy, W6 (green, Fig. 4),

dominated by BFl showed no clear tuning across direction for subject S010 but

may have played a role in the control of body position during the pPA period,

as the other muscles were activated for movement preparation.

112

Figure 5. 4 A: Muscle synergy vectors (W) and B: recruitment coefficients (C) for a

representative subject (S010) across all 5 time bins of the pPA. Synergy activation coefficients

for individual trials are shown by a dot, and average muscle synergy recruitment is shown by a

solid line that illustrates its directional tuning.

113

5.5.3 Muscle synergies accurately predict muscle activity patterns

To validate the extracted synergies, EMG patterns from REM trials

were reconstructed from the synergies extracted from the CTRL trials. The

reconstruction of EMGs for a representative subject is shown in Figure 5.5. For

this particular subject, the linear summation of 6 muscle synergies could

account for the variation in muscle activity across temporal bins of the pPA and

reach direction. Mean muscle tuning curves demonstrate that the predicted

muscle activity (solid black line) and the recorded EMG (dashed line) for each

postural muscle was well represented, as indicated by the significant goodness

of fit values (Fig 5.5). The mean VAF across all muscles in the pPA5 period for

S010 was 88.84 ± 8.18 (SD). The contribution from each muscle synergy is

depicted with different colored lines. For most muscles, overall activity was

achieved with contributions from mainly from one or two synergies, which is

particularly evident in later bins pPA4 and pPA5.

Muscle activation patterns varied from trial to trial, even within a single

reaching direction, as reflected by the variability in the muscle tuning curves.

The differences in activation could be explained by changes in the activation

coefficients of the muscle synergies, rather than modifications in the muscle

synergy structures (Fig. 5.6). Figure 5.6 shows the reconstruction of pPA4 and

pPA5 of the 6 REM trials for 75° reaching movements for S010 using the

muscle synergies extracted from the CTRL trials (muscle synergies shown in

Fig 5.4). Histograms depict the overall predicted muscle activity of each muscle

as well as showing the relative contribution of each synergy to the muscle’s

activity. Comparison with the actual muscle activation amplitude (indicated by

the stars) shows good predictions by the muscle synergies. Mean VAF across

all trials for pPA4 and pPA5 was 92.19 (±0.78 SD) and 92.66 (±0.77 SD),

respectively.

114

Figure 5. 5 Reconstruction of mean muscle tuning curves using the muscle synergies shown in

Fig. 5.4 for a representative subject (S010). Dashed lines represents observed data and solid

lines represent reconstructed data. Each muscle synergy’s contribution is shown by the

corresponding colored line. Combined, these result in the total reconstruction. Goodness of fit

(VAF, r2) of the reconstruction to the observed EMG is indicated.

115

Figure 5. 6: Postural muscle activity in the REM trials for reaching to the 75° target is

reconstructed using the extracted muscle synergies for a representative subject (S010).

Variations in the activation levels of the muscles between trials is well accounted for by the

modulation of the muscle synergies. Muscle activation amplitudes for all recorded muscle are

grouped along the x-axis within a trial. Star (✴) represents the observed EMG and open circle (

○) is the reconstructed activation amplitude. The relative contributions of the muscle synergies to the overall muscle activity is shown by the colours in the vertical histogram for each muscle.

116

5.5.4 Comparison of muscle synergy structure between subjects

Across the subjects analyzed, the number of muscle synergies extracted

ranged from a minimum of 5 to a maximum of 8 muscle synergies across all 14

recorded muscles in the pPA period. A comparison of these subjects is shown

in Figure 5.7 and reveals significant variability in the composition of the muscle

synergies between subjects. Across the subjects presented here, only 2-3 muscle

synergies were shared between subjects (Fig. 5.7). Furthermore, 2 subjects

demonstrated completely unique patterns of muscle synergy organization.

117

Figure 5. 7 Muscle synergy structure compared between subjects. Muscle synergies that are

shared between subject are indicated by a significant VAF and r2. Muscle synergies whose

backgrounds are shaded gray are specific to that subject.

118

5.6 DISCUSSION

The primary objective of the present study was to perform a

factorization analysis upon a large set of postural muscles that were activated

using an entirely feedforward mode of postural control. By doing so, it

attempted to identify whether postural muscles were recruited as synergic

groups, as previously documented for feedback postural behaviours (Chvatal et

al. 2011; Safavynia and Ting 2011; Ting and Macpherson 2005; Torres-Oviedo

et al. 2006; Torres-Oviedo and Ting 2007; 2010). Using a well documented and

proven NNMF method (Chvatal et al. 2011; Lee and Seung 1999; Safavynia

and Ting 2011; Ting and Macpherson 2005; Torres-Oviedo et al. 2006; Torres-

Oviedo and Ting 2007; 2010; Tresch et al. 2006), the present study found that

5-8 muscles synergies could account for the spatial and temporal patterns of

muscle activity in the supporting limbs prior to the onset of reaching

movements in multiple directions in the horizontal plane. This study extends

the results of our previous one documenting directionally-tuned muscle activity

for feedforward postural control (Leonard et al. 2009) and suggests that the

CNS could rely upon shared motor modules for simplifying the complex task of

balance control.

5.6.1 Modular organization of feedforward postural adjustments

This study has quantitatively defined groups of muscles recruited

together to serve the coordination of preparatory postural activity preceding

multidirectional reaching movements performed in standing. The variability in

the EMG across reach directions and across trials within a single reach

direction was explained by changes in the synergy activation coefficients, rather

than changes in the synergy structures. This supports the hypothesis that the

CNS relies on a modular structure for organizing motor behaviour (Cheung et

al. 2005; Ting and Macpherson 2005), whereby a motor command defines the

spatial and temporal recruitment of a synergy (Drew et al. 2008; Ting 2007).

119

These findings are consistent with several studies that have shown evidence for

muscle synergies in a variety of tasks (Cappellini et al. 2006; Cheung et al.

2005; Chvatal et al. 2011; d'Avella and Bizzi 2005; Hart and Giszter 2004;

Ivanenko et al. 2005; Safavynia and Ting 2011; Ting and Macpherson 2005;

Tresch et al. 1999).

Our findings are in agreement with previous studies that have shown a

modular organization of postural muscle activity associated with voluntary

movements (Krishnamoorthy et al. 2003; Krishnamoorthy et al. 2004). In these

studies, 3 - 5 principle components (PC) were identified to reflect muscle

synergies, or M-modes, related to consistent directions of force production

(Krishnamoorthy et al. 2003) and recruited according to stability of the support

surface (Krishnamoorthy et al. 2004). Fewer synergistic groups of muscles

were found in these studies as compared to our results for multidirectional

reaching movements. Likely, the differences in the number of synergies

recruited were due to differences in the list of muscles recorded or the task

constraint. Here, we asked subjects to perform reaching movements in 13

directions, whereas in the studies of Krishnamoorthy and colleagues,

perturbations resulting from the movement were limited to the sagittal plane

(Krishnamoorthy et al. 2003; Krishnamoorthy et al. 2004). The methods used in

these studies do not extract muscle synergies in the same manner as those used

here, and therefore the differences in the number of extracted synergies could

be due to this, although there does not appear to be an advantage of using one

methodology over another (Tresch et al. 2006). We specifically chose NNMF in

order to draw comparisons with the organization of feedback-based postural

responses documented in the literature and examine whether similarities for

feedback and feedforward control, as discussed in the next section.

A question raised by our study is how the muscle synergies relate to the

biomechanical parameters of the feedforward postural adjustment, such as end-

point force or motion of the CoM. In balance control, functional muscle

synergies have been analyzed in relation to the end-point forces exerted at the

ground (Krishnamoorthy et al. 2003; Ting and Macpherson 2005; Torres-

120

Oviedo et al. 2006) as well as motion of the CoM (Chvatal et al. 2011). These

results, along with results from biomechanical modeling studies (Welch and

Ting 2008), suggest that the muscle synergies are modulated by global

variables of balance control related to stabilization of the CoM rather than local

joint changes (Chiel et al. 2009; Lockhart and Ting 2007). Whether this is also

the case for the control of balance during the execution of movements, where

both postural and movement goals must be coordinated, remains to be

determined. To assess the organization of pPAs within the context of a

hierarchical control scheme where CNS relies on muscle synergies to translate

task-level goals into the appropriate muscle activation pattern, an analysis of

functional synergies is required and represents a direction of future research.

5.6.2 Similar organization for feedforward and feedback postural control

The finding that muscle synergies can explain the spatial and temporal

patterns of feedforward postural adjustments extends our previous study that

suggested feedforward and feedback postural control may be recruited via

shared neural pathways in the CNS (Leonard et al. 2009). Here, we

demonstrated that variations in both the temporal and spatial characteristics of

pPAs were accounted for by the modulation of the synergy activation

coefficients, which are thought to reflect the neural command that modulates

the synergy (Ting 2007; Ting and Mckay 2007). This strategy has been

consistently demonstrated for feedback postural control (Ting and Macpherson

2005; Torres-Oviedo et al. 2006; Torres-Oviedo and Ting 2007). Moreover, the

muscle synergies for feedback control are related to endpoint force production

(Ting and Macpherson 2005) and CoM motion (Chvatal et al. 2011), and

remain stable when examined across various biomechanical contexts (Torres-

Oviedo and Ting 2010). Whether this is also the case for feedforward postural

adjustments remains to be investigated.

Our analysis has shown that a similar number of directionally-tuned

muscle synergies to those observed in feedback (Torres-Oviedo and Ting 2007;

2010) can explain the feedforward postural muscle activity for pPAs preceding

121

multidirectional reaching movements performed in standing. Also, feedforward

muscle synergies showed clear directional sensitivity for certain reach

directions. In particular, the synergies were spatially tuned to rightward,

forward or leftward directions. Although direct comparisons to the muscle

synergies extracted for APRs (Torres-Oviedo and Ting 2007; 2010) cannot be

made as different muscle groups were recorded, several characteristics

generalize between the two modes of postural control. In both pPAs and APRs,

the task involves recruiting postural muscles to accelerate the CoM within the

BoS. Given that a similar number of directionally-tuned muscle synergies are

activated for both tasks, it is possible that the CNS recruits the same motor

modules for controlling the CoM regardless of the mode of control. Indeed,

based on signals recorded in the pontomedullary reticular formation (PMRF)

for both feedback and feedforward control have been identified in the cat

(Schepens et al. 2008), it is proposed that this structure may be responsible for

organizing postural control across the two modes. Therefore, we propose that

the CNS might rely on shared structures in both feedback and feedforward

postural control and that this integration occurs in the PMRF of the brainstem

(Leonard et al. 2009; Schepens et al. 2008).

It is interesting to note that the composition of the muscle synergies for

feedforward postural control exhibited greater variability between subjects than

those observed for APRs, which were also extracted using NNMF (see Fig 7 in

Torres-Oviedo and Ting 2007). Generally, for feedback balance control tasks,

such as restoring perturbed balance, several (approximately 4) muscle synergies

are shared between human subjects and exhibit consistent similarities in the

tuning of their activation coefficients (Torres-Oviedo and Ting 2007; 2010),

suggesting that the modular organization reflects the underlying neural control

(Ting 2007). Here, we show instead that it was an exception for muscle synergy

structure to be shared across subjects. It is hypothesized that descending

commands from higher neural centers recruit muscle synergies (Roh et al.

2011), therefore, given the voluntary nature of the task, subjects may have

adopted different strategies to perform the task. In this present experiment, no

122

instruction regarding movement speed or pattern were given, therefore subjects

were free to select their own movement strategies. Therefore, it is possible that

subjects recruited functionally equivalent synergies that differed in their

composition due to redundancy of their respective biomechanics (McKay and

Ting 2008). Further study is needed to address these differences.

5.6.3 Conclusions

We have shown that subject-specific muscle synergies explain the

spatial and temporal coordination of postural muscle activity organized in a

feedforward manner preceding reaching movements in multiple directions. In

general, 5-8 muscle synergies were sufficient to reconstruct the feedforward

postural EMG for each subject. However, only a small number of synergies

were shared across subjects. It is unclear whether this is related to the voluntary

nature of the task or whether subjects recruit distinct synergies that may be

functionally equivalent due to redundancy of the musculoskeletal system

(McKay and Ting 2008). These results are consistent with published

observations that muscle synergies represent a modular organization of motor

control, although they fall short of explaining how the synergy recruitment can

explain the force constraint strategy documented in our previous study

(Leonard et al. 2009). Future studies will examine how the synergic muscle

activity relates to the biomechanics of feedforward postural control. While

these results show similar features in number of synergies and spatial

organization to those observed for APRs, robustness of muscle synergies for

feedforward posture control remains to be determined by evaluating whether

the same synergies are preserved for postural adjustments accompanying

reaching movements. Overall, these findings provide a first step in examining

whether a set of robust muscle synergies exists for feedforward and feedback

postural control.

123

5.7 ACKNOWLEDGEMENTS

We thank Dr. Seyed Safavynia (Wallace H. Coulter Dept. of

Biomechanical Engineering, Georgia Inst. of Technology and Emory Univ.) for

advice about data analysis.

124

125

Chapter 6

Postural adjustments for online corrections

of arm movements in standing humans

6.1 PREFACE

The previous two chapters examined the postural strategies and

coordination of muscle activity for feedforward postural adjustments preceding

and accompanying goal-directed reaching movements. In these studies, the final

goal of the movement was planned and the disturbances associated with the

displacement of the limb and body towards the target could be predicted.

Consequently, pPAs and aPAs could be planned in feedforward accordingly

(Bouisset and Zattara 1981; 1987). However, in situations where the target goal

may change once the movement is initiated, the planned postural activity is no

longer appropriate for the updated focal movement. It is not known whether the

CNS updates posture predictively, or whether it relies on feedback from the

moving limb to make the correction in the posture. Therefore, in Chapter 6, I

address this question by determining the nature of the aPA command in relation

to an online correction of reaching movement executed in standing. This is the

first study to specifically examine the online control of posture using a

paradigm of visual perturbations of reaching movements performed in standing.

This chapter was adapted from Leonard JA, Gritsenko V, Ouckama R

and Stapley P.J. Postural adjustments for online corrections of arm movements

in standing humans. Journal of Neurophysiology, 105(5): 2375-2388, 2011.

This manuscript has been reprinted with permission from The American

Physiological Society, publisher of the Journal of Neurophysiology. The paper

is presented in the same format in which it was published with the exception of

126

formatting to figures and tables to comply with McGill University thesis

formatting guidelines.

6.2 ABSTRACT

The aim of this study was to investigate how humans correct ongoing

arm movements while standing. Specifically, we sought to understand if the

postural adjustments in the legs required for online corrections of arm

movements are predictive or rely upon feedback from the moving limb. To

answer this question we measured online correction in arm and leg muscles

during pointing movements while standing. Nine healthy right-handed subjects

reached with their dominant arm to a visual target in front of them and aligned

with their midline. In some trials the position of the target would switch from

the central target to one of the other target locations 15°, 30° or 45° to the right

of the central (midline) target. For each target correction, we measured the time

at which arm kinematics, ground reaction forces, and arm and leg muscle EMG

significantly changed in response to the target displacement. Results show that

postural adjustments in the left leg preceded kinematic corrections in the limb.

The corrective postural muscle activity in the left leg consistently preceded the

corrective reaching muscle activity in the right arm. Our results demonstrate

that corrections of arm movements in response to target displacement during

stance are preceded by postural adjustments in the leg contralateral to the

direction of target shift. Furthermore, postural adjustments preceded both the

hand trajectory correction and the arm muscle activity responsible for it, which

suggests that the CNS does not depend upon feedback from the moving arm to

modify body posture during voluntary movement. Instead, postural adjustments

lead the online correction in the arm the same way they lead the initiation of

voluntary arm movements. This suggests that forward models for voluntary

movements executed during stance incorporate commands for posture that are

produced based on the required task demands.

127

6.3 INTRODUCTION

When standing humans reach out to point at or grasp an object, the

central nervous system (CNS) must resolve two major task constraints: the

production of the correct hand and arm trajectory towards the target and the

necessary associated postural adjustments (aPAs) in the supporting limbs and

trunk for the maintenance of equilibrium. Using prior knowledge of the

dynamics of the moving arm and the internal disturbances that arm movements

inflict upon the body, the CNS can anticipate the impending disturbance and

correctly program the aPAs accordingly. However, what happens when the

final position of a visual target unexpectedly changes after the onset of a

planned voluntary movement that is executed while standing? In this situation,

the CNS must correct online the arm trajectory toward the new target and

modify the required postural adjustments. The objective of this study is to

investigate how posture is modified with respect to arm movements during

visual perturbations of the reaching goal. The online control mechanisms for

arm movements have been extensively investigated during experiments with

seated subjects. These studies have shown that modifications of hand trajectory

in response to target displacements occur at short latencies of 100 to 150 ms

(Day and Lyon 2000; Goodale et al. 1986; Paulignan et al. 1990; Prablanc and

Martin 1992; Prablanc et al. 1986; Soechting and Lacquaniti 1983; Zelaznik et

al. 1983). Other studies have shown that visuomotor corrections are automatic

and occur without voluntary intervention (Day and Lyon 2000; Gritsenko et al.

2009). Because of the inherent delays associated with the use of sensory

feedback, the short latencies of online correction support the notion that the

CNS adopts predictive mechanisms to execute rapid arm movements to visual

targets. Such a prediction involves the formulation of an initial plan of the

movement using a feedforward mode of control, but also a continuous

estimation of the actual state of the system compared to the desired one, which

is achieved using rapid, internal feedback loops (Bhushan and Shadmehr 1999;

Desmurget and Grafton 2000; Mehta and Schaal 2002; Sabes 2000; Shadmehr

and Krakauer 2008). This efference copy-based state estimation ensures that

128

motor commands can be modulated at short latencies and online, automatic

corrections of movement be made without detriment to the resulting movement

(Nijhof 2003; Saunders and Knill 2003).

Anticipatory modulation of muscular activity in the supporting limbs

that precedes the onset of voluntary movements during stance also suggests that

a degree of prediction of the future state of the body occurs in advance of an

impending disturbance to posture (Davidson and Wolpert 2005). The

characteristics of aPAs occurring before the onset of movement depend on the

prior knowledge of arm and body dynamics (Bouisset and Zattara 1981; 1987).

In these paradigms (e.g., arm raising) however, a postural disturbance can be

predicted in advance, and need not be modified online during the execution of

the movement, as the end goal matches that for which the postural adjustments

were initially programmed. Indeed, during reaching to fixed targets in multiple

directions when standing, feedforward postural adjustments follow a consistent

spatial pattern both before and during the movements (Aruin and Latash

1995a). What happens however, when postural adjustments for reaching,

programmed based on an initial state and an expected outcome, must be

modified because of an unexpected change in the visual position of the final

goal? Does the CNS still adopt a predictive mode of control or does it correct

arm trajectory before posture, which is then updated based upon the feedback

obtained from the arm correction? We attempted to investigate these questions

by inducing unexpected shifts in the visual location of the target after the onset

of reaching movements during stance. All targets were placed at a distance such

that subjects could point to them and retain their center of mass (CoM) within

the support base, without the need for a corrective step. Thus, subjects were

aware that they could complete the corrections and not become unstable to the

extent that they would lose balance. Therefore, two mechanisms of postural

control were possible: 1) Postural adjustments occur after the arm movement

corrections for the visual perturbations, and the feedback from the change in

arm and body configuration is used to update the necessary postural

adjustments, or 2) Postural adjustments are predictive and precede online

129

corrections of arm movement. This may indicate that, rather than acting to

reduce the disturbance induced by the upcoming arm correction, the postural

control system participates in the movement component of the action. We

hypothesized that because a change in the trajectory of the arm in response to

an unexpected shift in target position could potentially destabilize one’s

balance, the CNS predicts the postural adjustments necessary to execute smooth

online corrections of arm movements. In other words, when the target shift

occurs, leg muscle activity is updated and precedes the necessary changes in

ground reaction force in advance of changes in arm muscle activity or in

trajectory toward the new target. We show that this is indeed what occurs for

online corrections of arm movements during stance.

6.4 METHODS

6.4.1 Subjects

Nine right-handed subjects (5 females, 4 males) were recruited from the

McGill University student population to participate in the study. Subjects had a

mean age of 22.9 ± 3.1 (SD) years and measured on average 1.68 ± 0.1 m and

62.4 ± 9.6 kg in height and weight, respectively. All subjects were free of any

known neurological, visual, or orthopedic disorders, and provided their

informed consent to participate in this study. The study had ethical approval

from the McGill University research ethics board.

6.4.2 Experimental apparatus and set up

Subjects stood barefoot on two triaxial force plates (model FP4060,

Bertec, Columbus OH) that recorded ground reaction forces (GRFs) and

moments in the mediolateral (x), anteroposterior (y) and vertical (z) axes at

1000 Hz. Each subject stood with their feet positioned according to their

preferred mediolateral stance width, which was based on the average distance

between the 2 heels calculated immediately after 3 trials of walking 15 m across

130

the laboratory. This preferred stance configuration was recorded and marked on

the force plates and subjects maintained it throughout the experiments.

The experimental apparatus consisted of 4 target lights arranged about a

semi-circular radius separated by 15° (see Fig. 6.1A). The apparatus was fully

adjustable to each subject’s reach distance and height. Light targets were 2.5 cm

in diameter and consisted of a 5-V red light-emitting diode (LED) encased in a

modified gaming switch (model 459512; RP Electronics, Burnaby, BC,

Canada) that produced a 5-V pulse upon contact. Targets were mounted on

lightweight aluminum dowels, adjustable in length, affixed to a semicircular

aluminum bar suspended from the ceiling. Subjects wore a chest band with the

same switch that enabled the detection of movement onset upon its release.

Targets were situated at a distance corresponding to 130% of each subject’s

reach length measured to each respective target. Previous experiments adopted

this distance, which was attained comfortably without the centre of pressure of

either foot leaving foot length or width (Leonard et al. 2009). Any trials where

this occurred were rejected from further analysis. The choice of 130% was

especially important as we sought to evoke postural adjustments for online

corrections readily distinguishable above those produced for ongoing (initial)

reaches to the central target. Reach length (100%) was measured as the distance

between the xiphoid process (where the finger tip was held at the start of each

trial) to the tip of their outstretched finger when the arm was extended in the

direction of each of the targets (neutral scapula retraction).

The muscle activity of 16 leg, trunk and arm muscles was sampled at

1000 Hz using two DelSys Bagnoli 8-channel systems (Delsys, Boston, MA).

For all subjects the activity of the following leg muscles was recorded

bilaterally: tibialis anterior (lTA and lrTA), soleus (lSol, rSol), peroneus longus

(lPerl, rPerl), rectus femoris (lRF, rRF), biceps femoris (lBF, rBF) and tensor

facia latae (lTFL, rTFL). Additionally, recordings of muscle activity of the

reaching arm (right) included anterior and posterior deltoid (rADelt and rPDelt,

respectively), long head of the triceps (rTric) and the long head of the bicep

muscle (rBic). Bilateral kinematics were collected using a 6-camera MX3

131

motion-capture system (Vicon Peak, Lake Forest, CA) sampled at 200 Hz. A

custom written program using LabVIEW (National Instruments, Austin, TX)

controlled the illumination of the target lights, and acquired and synchronized

signals from the chest and target switches. Synchronization with the analog

signals obtained from the force plates and EMG system was done using the

Vicon controller.

6.4.3 Experimental procedures

Subjects began each trial standing with the head forwards in the

direction of the central (90°) target, which was aligned along the midline of the

body (see Fig. 6.1A). Before each trial, they depressed the chest switch attached

at the xiphoid process with their right index finger. The left arm was held in a

relaxed downward pointing position at the side of the body. The study

comprised 2 principal types of trials: regular reaches (‘reach’ trials) to the

central (90°) target and trials that required corrections of arm trajectory towards

3 other targets at different times after a ‘reach’ was initiated (online corrections

or ‘corr’ trials).

Subjects were asked to stand quietly before each trial began. Once the

experimenter was satisfied that the subject was stable, he/she initiated data

collection and after a variable delay, the central target light (90° or ‘L1’) would

illuminate. Subjects were instructed to reach and press the illuminated light at

their natural speed (‘reach’ trials, dashed trajectory, Fig. 6.1A), remaining in

that position until told to return (approx. 2 secs). For some trials, the target light

would shift from L1 to any one of the other 3 targets, 75° (‘corr75’), 60°

(‘corr60’), or 45° (‘corr45’) at different times after the onset of the initial reach

movement. All ‘reach’ and ‘corr’ trials were randomly presented as well as

trials when subjects were prepared, but no light illuminated. For the ‘corr’

trials, subjects were instructed to correct arm trajectory when they detected the

light change and point to the newly illuminated target. The target shift could

occur after a variable delay from the online detection of a voltage drop that

occurred when subjects released the chest switch. Figure 6.1B illustrates a

132

representative timeline for a typical corr45 trial at 442 ms after movement

onset. The first light (L1) illuminated approximately 250 ms after the auditory

tone. The subject reacted to L1 onset by initiating a movement of the finger

(‘movt onset’ in Fig. 6.1B). At this time, chest switch voltage dropped to zero.

After 442 ms, L1 turned off and light 2 (L2) illuminated. In this example, the

corr45 movement resulted in a lengthening of movement time

(Balasubramaniam and Turvey) of 322 ms with respect to the mean ‘reach’

movement.

133

Figure 6. 1: Experimental set-up and data collection schema. A. Subjects stood on 2 force

plates reached to a central target, aligned with their xiphoid process. Unperturbed ‘reach’ trials

were interspersed with online correction (‘corr’) trials involving unexpected illuminations of 1

of 3 other targets placed successively at 15° increments to the right of centre. B. Explanation of

the changes in voltage related to the sequence of light changes. When the signal rose to 5V each

light was illuminated. L1 = light one, L2 =light 2, chest = chest switch attached around the

subjects sternum that acted as a signal from which L2 illumination could be triggered. C. A

histogram showing the distribution of L2 onset as a percentage of mean ‘reach’ peak velocity.

Trials from all ‘corr’ conditions have been pooled (n=652). rFin = right finger.

134

A variable delay existed between when the controlling Labview routine

received the switch signal (shift from 5 to 0V indicating movement onset) and

when it simultaneously extinguished L1 and illuminated L2. This was estimated

to be on average 62.1 (± 35.1) ms. Thus, even though standard delays of 200

and 350 ms between movement onset and light (L2) change were used, actual

delays of the light changes were calculated on a trial by trial basis, using the

rising edge of the L2 5V signal (see centre grey line on Fig. 6.1B). In order to

standardize when L2 changes occurred, delays were represented as a percentage

of mean ‘reach’ trial MTs. The distribution of L2 onsets as a percentage of

‘reach’ MT is shown in Fig. 6.1C. Even though light shifts occurred within the

acceleration phase of ‘reach’ trials (open bars), a higher percentage occurred

during the deceleration phase (grey bars).

Each experimental session began with subjects performing a series of

acclimatization trials, consisting of 5 regular reaches to each target in turn.

Following the acclimatization period, trials were organized in a random order.

We required subjects to execute a total of 60 ‘reach’ trials to the central target,

and at least 15 correction trials to each of the 3 other target positions. Fifteen

trials were given during which no target illuminated (catch trials). In some

subjects, more than the standard number of corr trials was collected to ensure

that a large enough database was established. Catch trials were included in an

attempt to prevent subjects from predicting and begin moving to the target

before L1 illuminated. All catch trials were eliminated from subsequent

analysis. For experimental conditions to have adequate controls, at least three

times the number of control trials compared to experimental trials must be

collected (Zar 1999). For our analysis the ‘reach’ trials were to act as controls

for detecting differences in the ‘corr’ trials. Thus, we aimed to collect at least

120 trials from each subject (1080 for all 9 subjects). A breakdown of the total

number of trials collected and retained after trial selection procedures (see Data

analysis) is given in Table 6.1. Subjects performed blocks of 40 trials between

which they would take 5-minute rest periods to reduce fatigue. Each data

collection period lasted 45-60 minutes.

135

6.4.4 Data analysis

Data analyses were performed offline using custom routines written in

Matlab (The MathWorks, Natick, MA). Kinematic data and GRFs were low-

pass filtered using a digital second-order Butterworth filter, with a 10Hz and

100Hz cutoff frequencies, respectively. Raw EMG signals were high-pass

filtered at 35 Hz, demeaned, rectified, and low-pass filtered at 10 Hz, using a

2nd-order Butterworth filter. All trials were visually inspected for stability of

Fz during the background period (500 ms of quiet stance before the first target

light illuminated). Any trials showing significant variation in Fz were

eliminated from further analysis.

136

Figure 6. 2: Determination of the online correction of finger trajectory (fcorrect). A. Plan view (x,y) of rFin average ‘reach’ trajectories +/- 1SD (dashed line with shaded grey area) in relation

to one ‘corr45’ trial (full black trajectory). Filled black circle is the onset of light 2 (L2 onset)

and the open circle is the time at which the corr45 x,y trajectory exceeded the average ‘reach’

trajectory plus 1SD for subject S6. B. Average (dashed line) plus 1SD of curvilinear rFin

velocity for a reach movement and one ‘corr45’ trial(full black line). Black vertical line is light

2 (L2) onset, grey vertical line is the time of online correction (fcorrect). Each corr condition

has been displaced rightwards and downwards for clarity, but the starting position was the same

for each. C, D: Explanation of how the correction of the EMG activity and GRFcorrect

associated with online corrections were determined. C. Calculation of EMGcorrect. The muscle

shown is the left soleus muscle, but the same procedure was used with all other muscles studied

(see Methods). The dashed trace and grey traces represent respectively, the mean ‘reach’ soleus

muscle activity ± 2SDs above and below the mean. The dark full trace represents the soleus muscle activity produced during an online correction movement, in this example a corr45

movement. The open circle indicates the time at which the corr45 soleus muscle activity

exceeded the mean+2SD ‘reach’ activity level. This time was taken as the EMGcorrect time

(for more detailed explanation, see Methods and Results). Abbreviations as previous figures,

except Movt end=movement time. D. Calculation of GRFcorrect. Method for determining

GRFcorrect is shown for the left shear force (Fx). The dashed trace and grey traces represent

respectively the mean ‘reach’ Fx and ± 2SDs above and below the mean. The dark full trace

represents the Fx exerted during an online correction movement (in this example corr45). The

open circle indicates the time at which the exerted force was significantly different from the

mean forces exerted in a ‘reach’ trial.

137

Trials were aligned to movement onset, which was determined on a

trial-by-trial basis to be the time at which the tangential velocity (derivative of x

and y velocity) of the right finger marker surpassed a value of 3% of the peak

velocity in that trial. Movement end was also established as when velocity

reduced below the 3% threshold in that trial. We, and others, have previously

used this threshold value to successfully determine movement onset and end

(Leonard et al. 2009; Shabbott and Sainburg 2009). Correction (‘corr’) trials

were also eliminated if: 1) they showed tangential velocity profiles that did not

illustrate a pronounced ‘trough’ indicating that a significant reacceleration

occurred. In other words, in these trials the subject may have hesitated

sufficiently to execute a reach directly to a ‘corr’ target, or 2) the trough

between the first and second peaks of tangential velocity (see Fig. 6.3B)

descended below the 3% initial peak velocity threshold, indicating that subjects

moved too fast and reached the central target before correcting to the ‘corr’

target.

Once these trial rejection procedures were implemented, a number of

essential measures were determined based on each subject’s average

trajectories. These included: fcorrect (the deviation of a ‘corr’ trajectory from

the mean of all regular reaches), EMGcorrect (the times at which particular

EMG traces during ‘corr’ trials significantly deviated from an average of

‘reach’ trials) and GRFcorrect (times at which forces in the 3 axes significantly

deviated from the average force exerted for ‘reach’ trials). The online correction

of focal movement (fcorrect) was detected on a trial-by-trial basis for all ‘corr’

trials using the tangential velocity of the marker placed on the right index finger

with respect to the original target. First, the mean ±1SD of all ‘reach’ trial x-y

trajectories was computed for a subject (see dashed trajectory and shaded area,

respectively on Fig. 6.2A). Second, the tangential velocity of each ‘corr’ trial

was compared to the mean tangential velocity of all ‘reach’ trials and an

algorithm calculated the time at which the reacceleration of the finger for the

‘corr’ trials (corresponding to the online correction) exceeded 1SD of the mean

‘reach’ trials. This is shown in Fig. 6.2B (fcorrect), occurring for this particular

138

trial at 460 ms, 169 ms after the change in the target light (L2). Similar methods

were used to calculate EMGcorrect and GRFcorrect, but corrections were

detected using a threshold of mean ±2SDs, rather than 1SD for fcorrect

(kinematics). Mean ± 2SDs was found to be more robust for detecting

EMGcorrect and GRFcorrect than mean ± 1SD. An illustration of the

calculation of these variables is given in the relevant section of the Results (see

below). After a stable initial posture, ability to detect significant EMG

corrections in arm or leg muscles and other ‘corr’ trial rejection procedures

were implemented, a total of 652 trials were retained and used for further

analysis for the 9 subjects after the unstable trials were eliminated as well as

those in which the algorithm for EMG or force corrections could not detect a

significant change with respect to regular reaches. Table 6.1 lists the total

number of trials collected and retained in each reaching condition.

A custom-written algorithm detected the time at which the EMG

activity of the muscles identified as participating in the corrections in ‘corr’

trials exceeded that produced for ‘reach’ trials, and values for each ‘corr’ trial

were verified on a trial-by-trial basis. For example, Fig. 6.2C illustrates soleus

muscle activity for the same ‘corr45’ trial as in Figs 6.2A and B (finger

trajectory), and shows how the EMG activity of this muscle increases well

above the mean+2SD of the reach trials, and well before the correction of the

arm trajectory (fcorrect). EMGcorrect was computed for a total of the 3 leg

muscles in the left leg as well as the 3 identified in the right arm (lSol, lTA,

lPerl, rADel, rPDel and rTric) and additionally rBic. A similar algorithm was

used to detect when shear (Fx) and vertical (Fz) force also increased above the

force produced during a regular reach (Fig 6.2D).

6.4.5 Statistical analysis

To detect significant differences in total movement times and fcorrect

values between ‘reach’ and ‘corr’ conditions a one-way analysis of variance

(ANOVA) was used, with experimental condition as the single factor. The

statistical comparison between EMG correction times of arm and leg muscles

139

was done using ANOVA with 4 factors. First, for each individual trial, the

EMGcorrect values of each postural leg muscle were subtracted from the

EMGcorrect values of each arm muscle. Positive differences indicate that leg

muscles change after the target jump before arm muscles. Values (EMGcorrect

of leg muscles minus arm muscle EMGcorrect) were sorted for trials in which

the target jump occurred during the acceleration or deceleration phases of the

initial reach movement. This was done to investigate if the leg/arm muscle

corrections were influenced by the extra time and feedback that may have been

available when target jumps occurred in the acceleration phase. Then ANOVA

was applied to these differences to determine the main effects of the following

factors: Target (corr75, corr60, and corr45), Postural Muscle (lSol, lPerl, and

lTA), Subject (9 subjects), and Phase (acceleration or deceleration of the arm

during the target jump). Post-hoc comparisons were performed with Sidak-

Bonferroni correction for alpha (Abdi 2007). Further post-hoc comparisons

were done using linear regressions between the EMG correction times for arm

muscles deemed as contributing to the online kinematic corrections and fcorrect

values, as well as between postural and arm muscle EMG correction times. For

each type of linear regression we report slope (m, the amount of increase in Y

that accompanies one unit of increase in X) and the Y intercept (Yi), the point

conventionally chosen to define Y coordinates when X=0. Thus, for the

relationship between leg and arm muscle correction times, we sought to

investigate what latency the arm muscle corrected at when the postural muscle

was 0 (or vice versa).

6.5 RESULTS

6.5.1 Unperturbed reaching and characteristics of online corrections

Trials executed to the central 90° target (‘reach’) typically showed an

early phase, up to peak velocity, during which the trajectory curved slightly

rightwards from the midline. This is illustrated by the average trajectory (± 1SD

of the mean) for one representative subject in Fig. 6.3A. The acceleration phase

140

was followed by a longer deceleration phase during which the trajectory curved

inwards towards the target. Trials that required online corrections of trajectory

showed significant deviations (fcorrect) from the reach trajectory, represented

by filled circles on average trajectories to each of the 3 targets. This deviation

became more accentuated as corrections were required further towards the right

target (45°). However, subjects were able to successfully correct their

trajectory, attain the targets and remain balanced.

141

Figure 6. 3: Reaching movement kinematic characteristics and profiles of curvilinear velocity.

Shown are averages plus 1SD for all trials for subject 2 in each of the 4 conditions studied. A.

Plan view (x,y) kinematics of rFin trajectory for ‘reach’ trials and each of the correction

conditions. B. rFin curvlinear velocity also for all 4 conditions. fcorrect=kinematic correction of

finger trajectory, mvt end=end of the focal movement (reach and corrected movements).

142

The curvilinear velocity of the right finger (rFin) marker enabled a very

clear identification of the online correction of the pointing movements

(‘fcorrect’, filled circles). Average velocity profiles for reach, corr75, corr60

and corr45 are shown in Fig. 6.3B for subject S2 (corresponding to the average

trajectories, Fig. 6.3A). Typically, at fcorrect, the finger reaccelerated and

displayed a second peak in velocity before the arm decelerated to the target

(‘mvt end’). The online correction of trajectory (fcorrect) occurred at, on

average 178.1 ± 58.3 ms (corr75), 187.6 ± 58.4 ms (corr60) and 191 ± 48.1 ms

(corr45) after L2 onset (values shown on Fig 6.2B for subject S2 are absolute

values after the onset of the movement). Movement times and fcorrect values

for all ‘corr’ conditions are given in Table 6.1. Movement times increased

between each of the 4 conditions from 815.5 ms (reach) to 1294.6 ms (corr45)

and there was a significant main effect of condition between reach, corr75,

corr60, corr45 (F(3,772)=492, p=0.00). However, there was no significant

difference between the fcorrect times to any of the 3 conditions that required

online corrections (corr75, corr60 and corr45), indicating that, when the whole

range of perturbations was considered, the online corrections were constant

with respect to the onset of the initial reaching movement and L2 onset.

143

Table 6. 1 Breakdown of total number of trials collected and retained after

trial selection procedure.

Condition n Movement Time (ms) L2 onset (ms) fcorrect (ms)

reach 228 877.6 (176.3) n/a n/a corr75 158 1090.7 (141.4) 360.1 (127.4) 567.7 (103.8)

corr60 148 1237.7 (137.8) 348.5 (76.9) 563.5 (87.6)

corr45 118 1318.8 (269.0) 340.1 (91.4) 566.5 (90.1)

Average movement times, light 2 (L2) onset and time to correction of finger

trajectory (the latter 2 for the applicable conditions) for all experimental

conditions tested. Latencies are calculated relative to the onset of the initial

finger (reach) movement. N = Number of trials.

144

6.5.2 Corrective forces and electromyographic activity accompanying

online corrections of arm movements

Typical EMGs and GRFs produced for both a ‘reach’ and ‘corr45’ are

shown in Fig. 6.4. Four arm muscles are shown on the same time scale as 6

bilateral lower limb muscles and the corresponding Fx, Fy and Fz forces.

During a typical ‘reach’ (Fig. 6.4A), the arm movement was initiated by the

activation of the rADel, and also bilateral anticipatory TA activity and Sol

inhibition in the legs. The effect of this postural muscle activity at the

beginning of the movements was to produce a backward directed Fy and a

loading (Fz) of the right foot to induce a forward sway of the body to the target.

Towards the end of the movement, the posterior deltoid and, to a lesser extent

the triceps and biceps muscles, became active. This activation of arm muscles

to brake the arm movement at the target was accompanied by associated

postural adjustments in the extensor muscles of the legs, represented in Fig.

6.4A by bilateral Sol muscle activity, starting approximately 500 ms after

movement onset. Forces showed that during this phase of the movement both

feet exerted force forwards (Fy) and the right foot Fz was loaded. This pattern

of EMG and force activity preceding and during the reach movements was the

same as that described previously (Leonard et al. 2009).

Movements necessitating online corrections of finger trajectory evoked

specific adjustments of both arm and leg muscle activity preceding fcorrect.

The adjustments in the postural muscles were recorded principally in the

muscles of the left limb during the period between light 2 onset (L2) onset and

fcorrect (shaded areas of EMG traces in greyed bar, Fig. 6.4B) and produced

distinct changes in Fx and Fz components of GRF under the left foot, as

compared to ‘reach’ trials (see vertical arrows). The most significant

adjustments in arm muscle activity were evident in the rADel, rPDel and rTric.

In the postural muscles, significant activations between L2 onset and fcorrect

occurred in the lSol, lTA and lPerl. During approximately the same period of

‘reach’ trials, no such postural or arm muscle activity could be seen (see Fig.

145

6.4A, ‘approx. area of fcorrect’). The postural adjustments from L2 to fcorrect

recorded during online corrections led to an increase in leftward-directed shear

force (Fx), and a reloading of Fz also under the left foot. Analysis of the times

to correction of arm and leg muscles corresponding to the online correction of

movement, and their relationship was therefore principally limited to this subset

of left side postural muscles (lSol, lTA and lPerl) and the 3 right arm muscles

(rADel, rPDel and rTric).

146

Figure 6. 4 Typical arm and leg muscle activity in relation to the 3D ground reaction forces

produced for a ‘reach’ movement (A) and an online correction movement to the target placed

45° to the right of midline, ‘corr45’ (B). In each, the muscles plotted in grey are those recorded

in the right leg. The vertical dashed line indicates initial movement onset (‘Mvt Onset’) and the full vertical black line, movement end (‘Mvt End’). In B., the shaded grey area indicates the

area in which arm and postural adjustments occurred. For muscle abbreviations, see Methods.

L2 onset=light 2 onset, fcorrect=time of kinematic correction of the rFin marker.

Fx=mediolateral force, Fy=anteriorposterior force and Fz=vertical force.

147

6.5.3 Arm-muscle activity responsible for corrections of finger trajectory

Of the right arm muscles recorded, we investigated which were related

to the online correction of finger trajectory. Figure 6.5 shows linear correlations

calculated between the latency of onset of fcorrect and EMGcorrect of each of

the recorded arm muscles (rPDel, rTric, rADel and rBic). Of these 4 muscles, it

can be seen that 3 (rPDel, rTric and rADel) showed highly significant

correlations between the onset of the corrective muscle activity of the arm and

the correction of curvilinear finger trajectory (fcorrect). The corrective muscle

activity in these 3 muscles preceded fcorrect by average values of -104.5 ms

(rPDel), -101.8 ms (rTric) and -30.4 ms (rADel), as determined by the intercept

of the regression lines. Therefore, based upon this, we sought to determine if

the left leg postural activity evoking changes in shear (Fx) and vertical force

(Fz) that reoriented the body towards the new target during online corrections,

preceded or not the corrective arm muscle activity in each of the 3 right arm

muscles that were correlated to fcorrect (rPDel, rTric and rADel).

6.5.4 Corrective postural adjustments in leg muscles lead arm muscle

corrections during online corrections of arm trajectory to

unexpected shifts of target position

Most of the EMGcorrect times in the postural leg muscles identified as

participating in the online corrections during ‘corr’ trials (lSol, lPerl and lTA)

were shorter than those in the reaching arm muscles related to fcorrect (rPDel,

rADel and rTric; Fig. 6.5). The average GRFcorrect values for Fx preceded

fcorrect by approximately 80-85 ms (corr45: 84.8 ms ± 40.6; corr60: 79.7 ms ±

85.5; corr75: -85.9 ms ± 69.6). Thus, it was likely that the postural activity

occurred before the changes in GRF and corrections of arm muscular activity

(see Fig. 6.4) and kinematics (fcorrect). ANOVA found a significant main

effect of the Subject factor (F = 26.23, p < 0.001), but not-significant main

effects of Target (F = 0.49, p = 0.61), Phase (F = 1.61, p = 0.20), and Postural

Muscle (F = 2.34, p = 0.10) factors. Post-hoc multiple comparisons show that

148

the EMGcorrect times in the postural muscles were significantly shorter than

those in the reaching muscles for the corr60 target (Fig. 6.6A). However, this

effect is significant for the lSol and lPerl muscles, but not for the lTA muscle

(Fig. 6.6B). Furthermore, the differences between postural and reaching

EMGcorrect times are even stronger for the corr45 target, in which the visual

perturbation was the largest and the inter-trial variability of the EMGcorrect

times was the lowest (Fig. 6.6B). However, the variability of differences

between the postural and reaching EMGcorrect times was large across subjects,

with 2 of 9 subjects showing shorter EMGcorrect times for the arm muscles

(Fig. 6.6C). Lastly, the differences between postural and reaching EMGcorrect

times did not vary between the acceleration and deceleration phases of the reach

(Fig. 6.6D).

Each of the EMGcorrect times in the postural muscles were linearly

correlated with the EMGcorrect times in the reaching muscles. All EMGcorrect

times were calculated with respect to the onset of the initial reach movement

across the entire range of L2 light onset latencies (Fig. 6.1C). Figure 6.7 shows

an example of 9 linear regressions between postural and reaching muscles (data

for all ‘corr’ trials were pooled) for a representative subject (S9), The Y

intercepts (Yi) of the regression lines across most subjects and most conditions

were largely positive, which indicates that most of the corrective postural

muscle activity started before the onset of the corrective reaching muscle

activity (Table 6.2). The advance activation of the leg muscles with respect to

arm muscles ranged from 16.9 ms (lPerl before rPDel) to 390.8 ms (lPerl before

rTric). The slopes (m) of each of the regression lines ranged from 0.256 (lPerl

vs. rPDel) to 1.71 (lSol vs. rTric).

149

Figure 6. 5: Linear regressions calculated between the four arm muscles recorded in the right

am and the fcorrect latencies calculated using the curvilinear kinematics of the rFin marker. A.

right posteior deltoid, B. right triceps, C. right anterior deltoid and D. right biceps. Yi=the value

of the Y intercept when X is zero.

150

Figure 6. 6: Multiple comparisons of differences between EMGcorrect values of arm and leg

muscles. A. Average differences EMGcorrect differences per ‘corr’ target. Values for all 3 arm

and leg muscles have been pooled (averages for each ‘corr’ target +/- 95% confidence interval,

CI). Positive differences indicate postural muscle corrections before arm corrections (see direction of arrow, top right of figure). B. Average differences (+/- 810 95% CI) EMGcorrect

(all arm muscles pooled) per leg muscle. Values for EMGcorrect measures were pooled for all

arm muscles and expressed as differences with each leg muscle in turn (positive differences

also indicate postural muscle corrections before arm muscles). C. EMGcorrect differences (+/-

95% CI) per phase of reach, i.e. before peak velocity (Acceleration) or after peak velocity

(Deceleration). Filled circles show mean EMGcorrect differences for all targets, while open

circles show only data for corr45; shaded area represents 95% confidence interval. D. Average

differences (+/- 95% CI) EMGcorrect (all arm muscles pooled and all leg muscles pooled) per

subject.

151

Figure 6.7 Linear regressions calculated between postural and arm muscles participating in the

online corrections to unexpected visual perturbations of target position. Each graph shows the

postural muscle correction latencies (EMGcorrect) as the dependent variable (A. lSol, B. lPerl

and C. lTA) with the regressions performed between the 3 arm muscle correction latencies

(rPDel, rADel and rTric). Times (in ms) are expressed along each axis from the onset of the

initial reach movement.

152

Table 6. 2 Slope (m), Y intercept (Yi), r2 values, p-value for the strength

of the regression fit (p) and p-value for the intercept (p-int) for linear regressions conducted between the leg and arm muscles selected to characterize the online corrections to all targets for each subject

* and ** = significant at p < 0.05 and < 0.01 respectively. n/a = no significant

modulation of EMG between corr and reach conditions. = Multiple comparisons of intercepts

between subject are significant when p-int is alpha of 0.0057 (Sidak-Bonferroni correction for 9 tests between muscles per subject).

153

When data was pooled for each ‘corr’ condition the activation of the leg

muscles consistently preceded those of the arm (positive Y intercept) and the

leg muscles evolved earlier during the corrections than arm muscle activation

(slope analysis). In order to verify if the predictive activation of leg vs. arm

muscles held across target shifts of different amplitudes (eccentricity), we

performed the same linear regression analysis for the pooled ‘corr’ conditions

for each subject. Table 6.2 provides a complete breakdown of these

relationships. Of the 49 linear regressions performed (3 x 3 muscles for 9

subjects), all but 9 showed significant r2 values. Importantly however, only 5

revealed that arm muscle activation preceded leg muscle activation (Yi values <

0 and m values >1). For most comparisons, values of slope were <1, and as low

as 0.107 (rADel vs. lPerl for S6), confirming, as with the pooled data, that the

activation of postural muscles evolved more rapidly than arm muscles.

Therefore, when all conditions and muscles were considered, our results

supported a predictive control of postural activity in relation to arm muscle

activity across all target directions.

6.6 DISCUSSION

This study investigated how the two components of posture and

voluntary movement interact when changes in ongoing reaching movements are

produced while standing. We proposed two possibilities: 1) when faced with an

unexpected change in visual target position, for which a reaching movement

had been initiated, postural adjustments would occur after the adjustments seen

in the arm muscles correcting the trajectory of the hand towards the target; 2)

postural adjustments would occur before any change in arm muscle correction

and kinematic change of hand position, and posture would therefore be

predictive with respect to the voluntary component of the action. Our results

supported the latter of these two possibilities, i.e. modifications in leg muscle

activity preceded those of arm muscles when target position changed,

regardless of how far a correction was required to be made with respect to the

midline (eccentricity). We will begin by discussing some methodological

154

differences between our paradigm and previous visuomotor corrections

paradigms. Then we will highlight the functional role of the online postural

adjustments, situate our kinematic correction results within existing literature

taken from double-step experiments carried out in seated subjects, and finish by

placing our findings within a theoretical framework of predictive motor control.

6.6.1 Methodological considerations

In this study, we used a digital signal, derived from a switch detecting

the onset of right finger movement, to trigger the change in light target from the

initial central one to one of the other 3 targets. Different delays resulted

between when the finger began to move and when the first and second lights

extinguished and illuminated, respectively. This paradigm was intended to

provide us with a means of investigating how humans adapt posture and

movement when programmed actions must be modified ‘online’ after they have

begun. Our paradigm cannot however, be regarded as a classical ‘double-step’

paradigm, one that involves a change in gaze saccade after an initial one has

been initiated, a paradigm traditionally been used to probe the properties of the

oculomotor system (Becker and Jürgens 1979; Westheimer 1954; Wheeless et

al. 1966). This is principally because our light target changes (from L1 to L2)

were not triggered using the onset of the initial gaze shift, as in other arm

movement studies (Goodale et al. 1986; Gritsenko et al. 2009). Nevertheless,

our paradigm can be likened to a number of arm movement studies that have

used a double step paradigm triggered on the onset of movement, either a 1-step

(amplitude), double-step (Gielen et al. 1984; Megaw 1974), or a 2-step

(amplitude and direction) one (Day and Lyon 2000; Georgopoulos et al. 1981;

Soechting and Lacquaniti 1983). For the purposes of this study however, using

delays from the onset of finger movement provided us with a window of time

sufficiently long in which to investigate changes in arm movements (and the

EMG activity related to them), and their associated postural adjustments. Given

that our L2 latencies averaged between 365-394 ms, and fcorrect latencies

between 556-565 ms, we were sure that we had a window of time (± 200 ms)

155

long enough to perform our analysis of both arm and postural muscles.

Moreover, during this period (and beyond) our focus was on the relationship

between arm and postural muscle activity to produce the corrections.

6.6.2 Postural adjustments contribute to the execution of voluntary

movement

An interesting aspect of our results, with regard to the postural

corrections associated with the change in hand trajectory, was that the postural

muscles showed activity that effectively increased the shear force under the

limb contralateral to the reaching arm prior to fcorrect. The vertical force under

this limb also showed loading during the same period. This would suggest that

the postural corrections acted to push out and down with the left foot, thus

helping to rotate the body rightwards towards the new target (see shaded area,

Fig. 6.4B). With this in mind, the postural corrections can be described as being

a component of the voluntary movement, rather than ensuring only the

maintenance of equilibrium. In other words, they accelerated the body in the

direction of the target and did not stabilize posture or compensate for the

impending arm correction. Had the latter been the primary objective of the

postural adjustments produced between L2 and fcorrect, significant activation

of right limb muscles would have been recorded to counter the impending

rightward rotation of the body when the arm was oriented towards the new

target. In light of this, it would seem difficult to divide so-called ‘posture’ and

‘movement’ components of the motor act, as has often been the case during

voluntary movement studies (Cordo and Nashner 1982; Hess 1954; Saltzman

1979). Rather, our results corroborate earlier work suggesting an integration of

postural and focal commands at higher levels of the CNS (Aruin and Latash

1995b; Commissaris et al. 2001; Stapley et al. 1999). Our results show that

postural adjustments contribute to focal corrections of voluntary movement.

156

6.6.3 Effects of standing on the characteristics of online corrections of the

arm

An interesting question that arises when examining the online correction

of arm movements during stance, is to what extent does upright posture

influence the time taken to initiate a correction once target position changes? It

is out of the scope of this study to make definitive conclusions with regard to

this question from an experimental perspective, as 1) we did not directly

compare online corrections in seated and standing conditions, and 2) in order to

obtain significant postural adjustments, the amplitudes of target shift used by us

far exceeded those adopted in seated studies. Nevertheless, comparisons can be

made between our study and previous ones in terms of the percentage of total

MT taken to initiate a corrective response, even though the mode of target shift

should also be accounted for (see above).

A detailed description of the early kinematic response and its relation to

the amplitude of target displacement has recently been reported by Gritsenko et

al. (2009). They used target shifts of 15 cm amplitudes in 8 possible directions

from a central one. Light changes were triggered during a period of saccadic

suppression of the initial saccade to the first light, and were estimated to occur

on average 50 ms before the onset of initial hand movements, which were

executed at preferred speeds (as in our study). For perturbed movements that

were on average 403 ms in duration, Gritsenko et al. (2009) reported that

corrections occurred at 35% (average 140 ms) of total MT. Longer ranges of

38-61% have however been reported by Prablanc and Martin (1992). Our

correction values occurred at, on average 47% of total MT across the 3 target

positions. Thus, it is likely that the mode of target shift trigger (gaze vs. first

hand movement) plays a role in the onset of the online correction. Despite this,

our results demonstrate that standing does not impede the early onset of the

correction, which is to some extent comparable to online corrections produced

when humans are seated.

Nevertheless, it is evident that both the total MTs and times to online

correction (fcorrect) reported by us were far slower (almost double) than those

157

reported during seated movements (Goodale et al. 1986; Gritsenko et al. 2009;

Komilis et al. 1993; Pélisson et al. 1986; Prablanc and Martin 1992; Sarlegna et

al. 2003). Previous studies using the double-step paradigm to perturb arm

movements during stance have reported correction times as low as 164-168 ms

(Fautrelle et al. 2010). However, these authors used targets changes that

required corrections in the sagittal plane only and triggered target position

change before or just after (50 ms) hand movement onset. To our knowledge,

the only other study of double-step perturbations during stance investigated

differences in reaction time or MT with the likelihood of a double-step

perturbation (Martin et al. 2000). Other recent studies have investigated online

corrections of the foot during walking or a single step (Reynolds and Day

2005a; b). The foot online correction onset has been reported to be between 239

and 402 ms after the foot off the ground for a step (Reynolds and Day 2005b),

while during swing phase of locomotion the foot online correction onset has

been reported to be more similar to that for the arm, 114-151 ms (Reynolds and

Day 2005a). This suggests that be delayed onset of foot online correction

during a step and, possibly, the delayed onsets of online correction observed in

this study are due to the increased information processing of equilibrium

constraints, which may be simplified during a predictive shift of CoM thought

to occur during locomotion (Day et al. 1997). Furthermore, Hollands et al.

(Hollands et al. 2004) showed that during tasks involving body rotation when

standing the onset of saccadic eye movements to targets is also delayed,

possibly by additional information processing of equilibrium constraints.

Interestingly however, taken together with studies of arm correction in seated

subjects and foot correction of stepping subjects, our results provide evidence

for a constancy of times to online corrections of arm movements regardless of

postural configuration. Both Gritsenko et al. (2009) and Prablanc and Martin

(1992) reported no significant differences in the time to correction across

different target amplitudes, and our fcorrect data also showed no significant

differences across the 3 ‘corr’ conditions.

158

It may be debated that the outward (rightward) curvature (rather than a

straight trajectory towards the central target) seen for the ‘reach’ trials was

influenced by the expectation of a change in target position. It is outside the

scope of the present study to definitively rule out if expectation was the cause

or not, of that curvature. However, it is interesting to note that trajectories

displayed for reaches to the same central target during a paradigm of

multidirectional reaching (Leonard et al. 2009) also displayed a similar

curvature (unpublished observations). Moreover, in a seated online correction

task, Day and Lyon (2000) also showed similar curvatures of straight reaches.

However, further study is warranted to investigate the role of target change

expectation in trajectory formation, in both the seated and standing positions.

6.6.4 Implications for the control of posture and movement

Skilled voluntary movement, such as reaching with the arm, relies upon

the prediction of the future state of the motor system because of the inherent

delays associated with information processing. Studies of anticipatory

adjustments of grip force with load during object manipulation (Flanagan and

Wing 1997; Johansson and Cole 1992; Kawato 1999) have shown that the CNS

‘predicts’ the dynamic effects of upcoming movements. Interestingly, such a

grip force/load force predictive relationship is preserved and precedes arm

movement online corrections, which is strong evidence of the use of predictive

motor mechanisms by the brain (Danion and Sarlegna 2007). Other studies of

the initiation of stepping suggest that the CoM is controlled predictively by the

CNS to maximise the efficiency of movement (Day and Lyon 2000; Day et al.

1997). These studies, and others that have examined the execution or learning

of seated arm movements, have proposed that forward, or predictive, internal

models are employed to anticipate the consequences of actions based on

efference copy of outgoing motor commands (Wolpert and Miall 1996).

Efference copy is used to estimate the sensory feedback likely to result from the

motor command, which results in accurate predictions of current motion

(Davidson and Wolpert 2005; Desmurget and Grafton 2000).

159

The use of such predictive models for controlling equilibrium during

arm movements has also been documented. The specific characteristics of

anticipatory postural adjustments occurring before the onset of movement

(APAs) depend on the prior knowledge of arm and body dynamics (Bouisset

and Zattara 1981; 1987). However, APAs are predictive with respect to the

initial onset of movement and are programmed with the CNS anticipating the

consequences of the impending act. Our results have shown that when the

initial goal of a programmed reach movement unexpectedly changes after it has

been initiated, the postural adjustments required to execute the online

corrections consistently lead any changes in arm EMG or kinematics (Figs.

6.6,6.7). Moreover, our results also show that, even for small amplitude

changes in target, requiring smaller corrections of hand trajectory and posture, a

predictive mode of postural control is still largely adopted (leg muscle

corrections led those of the arm consistently across ‘corr’ conditions, Fig. 6.7).

Therefore, we suggest that postural adjustments in this situation are largely

predictive in nature and not based upon feedback from the moving limb. Our

results corroborate those described by Ruget et al (2008) who showed that

human subjects are able to modify components of weight shift online during the

anticipatory period preceding the onset of stepping.

How do our findings fit with what is known about the neural control of

arm movements and posture as well as forward models of reaching movements?

It is known that there is a significant cortical involvement in predictive postural

behaviour. Mackinnon et al. (2007) demonstrated a facilitation of the muscles

involved in APAs whereas anticipatory adjustments of arm muscles were

absent in patients with damage to their motor cortex (Viallet et al. 1992).

Moreover, animal studies have suggested that the cortex contributes to motor

planning of reaching during stance (Martin and Ghez 1985; Perfiliev 2005;

Perfiliev 1998; Vicario et al. 1983) and the feedforward adjustments

accompanying the reach (Yakovenko and Drew 2009). Thus, the CNS may

send a global command that specifies the planning and execution of movement

and posture as one. Even though evidence has been given in support of global,

160

hierarchical signal in which pathways for movement give off collaterals for

posture (Gahery and Nieoullon 1978), the findings of a number of studies

would suggest that parallel, independent commands exist for the postural

adjustments preceding the movement, accompanying the movement, and indeed

the movement itself (Horak et al. 1984; Schepens and Drew 2004; Schepens et

al. 2008).

Although our results do not allow us to elucidate the origin and specific

anatomical structure of the pathways involved in posture and movement, we

can speculate about what happens, in terms of the production of postural

adjustments in relation to arm muscle activity, when the expected outcome of a

reach does not materialize, such as when target location unexpectedly changes

after movement onset. Figure 6.8 proposes a simplified model of how

commands movement and posture would fit with a forward model of arm

movements. The model shows that the motor cortex sends a global planning

command for movement and posture (Gritsenko et al. 2009). Efference copy of

commands for the execution of both the arm movement, APAs for initiating the

reach and aPAs accompanying the movement (right side of figure) ensures that

discrepancies of the expected movement are detected (Desmurget and Grafton

2000). All the time the expected reach movement mirrors that actually being

executed (above the grey horizontal bar), rapid feedback loops adjust and refine

the movement in real-time. When the target shifts, the expected movement no

longer reflects that which must be produced and a delay ensues. It is here that

adjustments of accompanying postural commands (aPA adjustments) must

occur before those of the arm (arm adjustments), in order that the dynamic

constraints of the task to be satisfied. Once posture is updated, predictively of

the arm movement, the online correction can be made and the target attained.

161

Figure 6.8 Suggested control schema illustrating the interaction of movement and posture

during online corrections of arm movements. APA = anticipatory postural adjustment, aPA =

associated postural adjustment (see text for explanation).

162

6.6.5 Conclusions

We have shown that when online corrections of ongoing arm

movements are required, the CNS adopts a predictive mode of postural control,

rather than a purely feedback-based mode. This was shown by the adjustments

of postural muscle activity consistently leading those of the arm muscles

responsible for correcting the trajectory of the arm to the target (Figs. 6.6,6.7).

Had the CNS relied upon information from a change in arm movement to

update posture, adjustments of arm muscle activity would have led those of the

postural muscles. This was not seen even for the smallest amplitude target

corrections (15° to the right), which could have been executed without

significant threats to stability.

6.7 ACKNOWLEDGEMENTS

This study was supported by a grant from the Canadian Foundation for

Innovation and a Natural Sciences and Engineering Research Council (NSERC)

of Canada discovery grant to PJS. The authors thank Prof. Robert Kearney

(Dept. Biomedical Engineering, McGill University) for his advice about data

analysis.

163

Chapter 7

General Discussion

Movements executed while standing incur a disturbance to equilibrium

due to the complex architecture of the skeletal and neuromuscular systems of

land-dwelling animals. For example, reaching to a target located just beyond

reach while standing (Flanders et al. 1999) requires the coordination of arm,

trunk, and supporting limb segment muscles to attain the visual target. It has

been shown that, in this instance the CNS predicts the dynamic consequences of

the motion, and plans postural adjustments preceding and accompanying the

movements to stabilize the body and assist in the performance of the movement

(Massion 1992). Presumably, coordinating these two aspects of a motor

behaviour is complex, and requires coordinated signals for the postural and

movement systems. The central organization of these two components of motor

behaviour remain elusive, therefore, the primary aim of this thesis was to

investigate the patterns of postural activity during goal-directed reaching

movements performed while standing.

To better understand how posture is controlled in relation to goal

directed movements, three studies were performed. The first study (Chapter 4),

examined how feedforward postural adjustments are organized spatially with

respect to movement direction. It demonstrated the existence of a force-

constraint strategy and muscle activity that is directionally tuned for pPAs. In

contrast, aPAs, while tuned in their EMG patterns, were characterized by a

greater dispersion of the direction of the GRF. With the goal of building on the

findings of the first study, the second study (Chapter 5) sought to determine

whether muscle synergies can explain the coordination of muscle activity for

feedforward postural adjustments preceding reaching movements. Results

showed that the coordination of muscle activity in the pPA period can be

explained by a time-varying recruitment of a few synergies. Finally, the third

164

study (Chapter 6) explored the nature of postural control signal associated with

online corrections of reaching movements. It showed that the CNS updates

commands to the postural muscles in advance of corrections to arm muscles

and finger trajectory.

Overall, these findings advance our knowledge of feedforward

(predictive) postural control. In particular, Chapters 4 and 5 showed that certain

characteristics of feedforward postural adjustments are similar to those

observed for feedback postural control. Furthermore, Chapter 6 demonstrated

that posture is prioritized and updated predictively when online corrections to

movement are required. Finally, this body of work provides baseline measures

of feedforward postural control as a first step to understanding why deficits in

balance control during voluntary movement may occur in the elderly, and other

pathological conditions.

In the following discussion, I will first review the significance of the

findings of tuned muscle activity, force constraint and muscle synergies for

feedforward postural control in relation to the strategies observed in feedback

postural control. Then, I will discuss how the findings can be used to interpret

possible modes of central control of posture and movement. Subsequently, I

will address the online control of posture within a broader scope of predictive

motor control. I will also review the relevance of these studies to the study of

balance in clinical populations. Finally, I will conclude by discussing some of

the unanswered questions raised by this thesis and directions of future research.

7.1 Characterization of feedforward postural adjustments during

multi-directional reaching movements

7.1.1 Role feedforward postural activity

Historically, the study of posture control during voluntary movement

has focused on the role of feedforward postural adjustments with respect to the

focal movement to gain insight into how these two components of motor

behaviour may be controlled by the CNS. Early studies of APAs suggested that

165

the postural activity preceding arm movements function to stabilize the body

against the disturbance of the movement (Bouisset and Zattara 1981; 1987;

Crenna et al. 1987; Massion 1992). However, subsequent studies investigating

forward reaching movements involving the trunk (Commissaris et al. 2001;

Stapley et al. 1998; Stapley et al. 1999), or those requiring a change in the

configuration of the BoS (Brenière and Do 1987; Brenière and Do 1986; Lepers

and Brenière 1995), have demonstrated that postural adjustments preceding the

movement onset instead create the dynamic forces required for initiating the

movement. In these situations, it appears that feedforward postural activity

contributes to the execution of the goal-directed component of the movement.

Similarly, pPAs and aPAs during unperturbed and perturbed reaching

movements contributed to the execution of the focal component of the

movement (Leonard et al. 2009; Leonard et al. 2011). For example, during

unperturbed reaching movements, pPAs and aPAs created the forces that

contributed to the initiation and termination of the reaching movement,

respectively (Leonard et al. 2009). In particular, a small forward displacement

of the CoM was observed prior to the onset of the reaching movement, as

defined by the onset of the finger movement. This was associated with

activation of the flexor muscles serving to initiate movement rather than

stabilize the position of the CoM. However, for the aPAs for unperturbed

reaching, strong extensor activity served to decelerate the CoM and stabilize the

position of the body towards the target. In the case of perturbed reaching

movements, the aPAs clearly contributed to overall movement correction by

breaking the forward motion and accelerating the body towards the new target

(Leonard et al. 2011). Together, these data support the view that posture

contributes to a voluntary movement by ensuring that the dynamic conditions

for the movement are met and suggest that task demands are integrated and

used to plan the appropriate postural activity using an internal model of posture.

166

7.1.2 Independent or parallel commands for global planning of posture and

movement

The fact that the postural adjustments preceding and accompanying the

movement contribute to the overall execution of the movement does not

necessarily indicate that these two commands are controlled by a single neural

command. Data showing that pPAs may be time-locked to the onset of a

voluntary movement have suggested that posture and movement are controlled

in a hierarchical fashion, where posture is controlled via collaterals from the

primary descending command for the movement (Massion 1992). However,

situations where the latency of the pPA is variable with respect to the

movement onset have also been observed (Brown and Frank 1987; Cordo and

Nashner 1982; Massion 1992), suggesting that instead posture and movement

are controlled via parallel, but independent neural pathways. There is also

debate concerning the nature of the signals for posture and movement for aPAs.

Until recently, it was generally accepted that postural commands are organized

in a hierarchical fashion in relation to the descending command for movement

via collaterals that recruit postural networks, likely localized to the brainstem

(Gahery and Nieoullon 1978; Massion 1992). This framework was largely

based on knowledge that cortical signals contribute significantly to the aPAs

(Mackinnon et al. 2007; Massion 1992; Viallet et al. 1992).

Recently, a series of studies of reaching movements in the cat have

provided considerable insight into the nature of the command signals for

posture and movement (Schepens and Drew 2006; 2004; 2003; Schepens et al.

2008; Yakovenko and Drew 2009). These studies identified independent signals

in the PMRF related to the pPA, the aPA, the movement, or the aPA and the

movement. What is clear therefore, is that a supraspinal structure, such as the

brainstem, plays a significant role in organizing both the postural adjustments

needed to execute reaching movements in the standing position, as well as the

voluntary movement itself. Interestingly, this same structure has also been

shown to control purely feedback-driven postural adjustments (Stapley and

Drew 2009).

167

The results of Chapter 6 demonstrated that the signal for the corrective

aPA consistently led the modulatory signal of the correction in the arm EMG

(see Fig 6.6 and 6.7). These findings are interpreted in the context of a forward

model of arm and posture (Gritsenko et al. 2009) in which the signals for

movement and posture are controlled via independent, parallel pathways

(Yakovenko and Drew 2009). In this framework, a global signal for posture and

movement is expressed, likely in premotor areas and sent to the motor cortex

(Massion 1992; Yakovenko and Drew 2009). Then, independent signals for the

pPA, aPA and movement are sent to the respective areas of the PMRF that

recruit and scale the activity of the appropriate muscle groups for the different

components of the motor behaviour (Schepens and Drew 2004). Meanwhile,

efference copies of these commands are consistently being evaluated and ensure

any discrepancies of the expected movement are detected (Leonard et al. 2011).

Accordingly, the corrective aPA must occur prior to the focal correction of the

movement and potentially inhibits the movement command until the conditions

for the movement have been met, as has been observed for pPAs and movement

initiation (Cordo and Nashner 1982; Massion 1992).

7.1.3 Strategies for simplifying the control of posture and movement

Several authors have argued that the CNS relies on neuromechanical

strategies to simplify the process of coordinating the many DoF associated with

complex task of controlling balance (Bernstein 1967; Chiel et al. 2009;

Macpherson 1991; Ting 2007; Ting and Mckay 2007). The postural strategies

have been examined in detail for reactive balance control using support surface

perturbations in cats (Macpherson 1988a; b; Ting and Macpherson 2005) and

humans (Carpenter et al. 1999; Henry et al. 2001; Horak and Nashner 1986;

Torres-Oviedo and Ting 2007). These studies have provided detailed analyses

of the spatial and temporal characterization of the muscle activity, force and

kinematic patterns for a range of perturbations spanning many directions, with

varying amplitudes and velocities. The resulting postural responses are

characterized by muscle activity that is directionally-tuned and GRF patterns

168

that are constrained. As well, the coordination of muscle activity across a large

number of muscles is explained by the temporal recruitment of a small number

of muscle synergies (Ting and Macpherson 2005; Torres-Oviedo and Ting

2007) according to task goal (Chvatal et al. 2011). Overall, these data have been

interpreted to support the hypothesis that the CNS relies on a hierarchical

organization to simplify movement control (Lockhart and Ting 2007; McKay

and Ting 2008; Ting 2007; Ting and Mckay 2007).

Prior to the work presented herein, no study had examined feedforward

postural control with the specific goal of drawing parallels to feedback-based

postural strategies. As such, Chapters 4 and 5 characterized the spatial

organization of the GRF and EMG patterns preceding (pPA) and accompanying

(aPA) reaching movements performed in standing across many directions of

reach with the goal of drawing comparisons to those cited in the literature for

APRs. Indeed, the characteristics of the muscle activity preceding reaching

movements (pPAs), i.e. directionally-tuned EMG explained by muscle

synergies and force constraint strategy, bear striking similarities to postural

responses organized in a feedback mode (Henry et al. 2001; 1998b;

Macpherson 1988a; b; Torres-Oviedo and Ting 2007). Interestingly, however,

the patterns of force in the aPA period exhibited greater dispersion yet still

displayed muscle activity that is sensitive to the direction of reach. The

significance of these observations will be discussed next.

7.1.4 Force constraint strategy: neural strategy or geometry?

Whether the force constraint strategy reflects a higher-level neural

strategy, or whether it is simply the outcome of the biomechanics of the limbs

requires some discussion. In studies of APRs, a characterization of the active

restoring force patterns revealed a clear bimodal relationship between direction

of CoM motion consequent to the perturbation and the direction of the GRF

(Henry et al. 1998a; 2001; Macpherson 1988a; Ting and Macpherson 2004).

Based on these findings, it was hypothesized that CNS constrains the direction

of GRF to simplify the control of motion by decreasing the DoF to be

169

controlled (Henry et al. 1998a; 2001; Macpherson 1988a). However, recent

modeling (Bunderson et al. 2010; McKay et al. 2007) and neurophysiological

(Honeycutt and Nichols 2010) studies have provided evidence that suggests the

force constraint may be in part due to the biomechanics and geometry of the

limb. It has also been shown that the CNS modulates the force generating

capabilities of the limb through the flexible recruitment of muscle synergies

(McKay and Ting 2008).

This framework may explain why aPAs were associated with forces that

were distributed relatively uniformly in relation to pointing direction in contrast

to the bimodal distribution observed in the pPA (see Fig 4.7 and Fig 4.8).

During the pPA period, subjects adopted an upright body configuration with

equal weight distributed between the left and right supporting limbs. However,

in the aPA, subjects leaned towards a single target with their right arm

extended. To maintain balance in this position, a greater contribution from a

single limb to the loading force was observed, as the CoM moves towards the

target and towards the limits of the BoS. Thus, it is possible that the orientation

of the limb and its musculature resulted in the distributed force patterns. This

would support the hypothesis that the biomechanics of the limb determine the

set of possible directions of applied force (Bunderson et al. 2010; Honeycutt

and Nichols 2010).

An alternative possibility is that the task goal may influence the patterns

of GRF directions. Modeling studies have shown that the space of possible

force production, which is inherently determined by the limb mechanics

(Bunderson et al. 2010; McKay et al. 2007), is actively constrained by the

recruitment of synergies (McKay and Ting 2008). As discussed in Chapter 4,

the muscle activity and forces generated in the pPA served to accelerate the

body forwards towards the target. In contrast, aPAs functioned to stabilize the

body at each target, which required precise control of the CoM and finger end-

point positions. The biomechanical constraints of this task may require

additional muscle synergies to those recruited for the pPA. Several studies have

shown the existence of ‘shared’ and ‘task-specific’ muscle synergies across

170

tasks (Chvatal et al. 2011; d'Avella and Bizzi 2005; Krishnamoorthy et al.

2004). Therefore, one could argue that when the goal is to accelerate the CoM,

as is the case in pPAs and APRs, the force constraint strategy may provide the

optimal solution. However, in situations where the CoM must be stabilized

uniquely with a different postural configuration, such as when the CoM

excursion approaches the limits of the BoS, additional task-specific muscle

synergies may need to be recruited, which may be related to additional

directions of force production. Further study is required to disambiguate these

relationships.

7.1.5 The importance of muscle tuning and synergic organization for

feedforward postural control

A salient finding of this thesis is that muscle activity was directionally-

tuned during both the pPA and aPA periods and a synergic organization of

muscle activity exists for the pPAs. Most postural muscles were modulated as a

function of reach direction, with maximal activity occurring for a small range of

directions (see Figure 4.5 and Figure 4.6). Within each period, several muscles

shared similar directions of maximal activity, suggesting that they may be

modulated together within a muscle synergy. Analysis of these relationships in

the pPA with NNMF revealed the existence of spatially-fixed muscle synergies

that were modulated as a function of reach direction (see Fig 5.4). These

findings contribute to our understanding of the central problems of motor

control by supporting current hypotheses of dimensional reduction for

simplifying motor control (Cheung et al. 2005; Ting 2007; Tresch et al. 1999).

Furthermore, the apparent similarity to the organization of feedback postural

control suggest that the CNS relies on shared structures to organize the muscle

activity for controlling balance regardless of the control mode.

Muscle tuning has been consistently observed in a variety of motor

tasks, including voluntary isometric contractions of the elbow joint (Buchanan

et al. 1986; Buchanan et al. 1989), feedback postural responses (Carpenter et al.

1999; Henry et al. 2001; 1998b; Honeycutt et al. 2009; Macpherson 1988b;

171

Ting and Macpherson 2004; Torres-Oviedo and Ting 2007), reactive stepping

(Chvatal et al. 2011), postural adjustments during voluntary movements (Aruin

and Latash 1995a), and goal-directed reaching (Flanders and Soechting 1990;

Scott et al. 1997; Sergio and Kalaska 1998; Thoroughman and Shadmehr 1999).

The observation of muscle tuning across such a range of behaviours suggests

that directional tuning is an inherent property of the motor system and largely

reflects the underlying anatomical and mechanical capabilities of a muscle in a

given orientation as a result of its complex architecture (Buchanan et al. 1986;

Herrmann and Flanders 1998). Therefore, the finding of tuned muscle activity

in Chapter 4 is not completely unexpected, and has been demonstrated in other

postural tasks associated with voluntary behaviour (Aruin and Latash 1995a).

However, inspection of the tuning curves in relation to one another can reveal

considerable insight into the contribution of individual muscles to the overall

torque generated and the inter-muscular coordination for different directions of

motor behaviour (Buchanan et al. 1986; Buchanan et al. 1989; Macpherson

1988b). Of interest is the finding that several muscles shared similar shapes of

their tuning curves and were activated together to subserve the postural activity

for a reaching direction (see Figure 4.6). This grouped tuning suggests that a

muscle synergy organization is used by the CNS to simplify the control of

feedforward posture by recruiting groups of muscles together as muscle

synergies. It has been proposed that this synergic organization is a mechanism

to simplify the control of movement by reducing the dimensionality of the

DoFs to control (Ting 2007).

Several studies have suggested that the CNS simplifies the control of the

musculoskeletal system by recruiting a small number of motor modules (Bizzi

et al. 2000; Cheung et al. 2005; Roh et al. 2011; Ting and Macpherson 2005;

Tresch et al. 1999). This has been shown for a number of tasks, including

reactive balance control in cats and humans, locomotor behaviours in the frog

(Cheung et al. 2005; Roh et al. 2011; Tresch et al. 1999), finger movements

(Shim et al. 2005), and balance control during voluntary whole-body sway

(Klous et al. 2012; Krishnamoorthy et al. 2003; Krishnamoorthy et al. 2004)

172

and arm movements (d'Avella et al. 2011; Muceli et al. 2010). In reactive

balance control, muscle synergies have been shown to modulate end-point force

and are thought to be a mechanism by which the CNS controls the CoM. These

relationships have been interpreted to suggest that the CNS recruits muscle

synergies to simplify the task of translating abstract task-level goals, such as

CoM kinematics, into the appropriate muscle activity (Cheung et al. 2005;

Chvatal et al. 2011; Ting and Mckay 2007; Torres-Oviedo and Ting 2010).

The analyses performed in Chapter 5 revealed that a muscle synergy

structure can reconstruct the temporal, spatial and inter-trial variability of pPAs,

which are organized strictly in feedforward (Massion 1992). Indeed, these

findings suggest that a modular organization of muscle activity, as has been

observed in feedback postural control, generalizes to the control of posture in

feedforward. However, in this study, only the muscle activity of the postural

muscles was evaluated without considering the relationship of the muscle

synergies to the forces exerted on the ground or the resulting CoM motion. In

order to make conclusions about whether the extracted muscle synergies are

related to task-level goals and whether a muscle synergy structure supports a

hierarchical organization of posture in feedforward, further study of the

functional muscle synergies is required. Furthermore, the robustness of the

muscle synergies remains to be determined by examining whether the

composition of the muscle synergies remains stable in different biomechanical

contexts, for example by varying stance width or during the aPAs. Nonetheless,

this thesis provides a first step by demonstrating the existence of muscle

synergies for postural adjustments preceding multi-directional reaching

movements. This provides further evidence that similarities exist in the central

control of posture for both feedback and feedforward modes of control,

suggesting that these signals are integrated within the CNS.

In summary, the synergy analysis performed in Chapter 5 shows

modularity as a basis for muscle coordination for pPAs. Together with the

findings of Chapter 4, these results suggest that feedforward postural control is

organized in a similar manner to feedback postural control. Unfortunately,

173

without analyzing the biomechanical variables with respect to the muscle

synergies, these data cannot specifically contribute to the discussion of a

hierarchical control of posture and movement in which task-level goals are

translated into muscle activation patterns. Instead, the similarities observed here

for pPAs and aPAs to reactive postural control strategies suggests that

commands for feedback and feedforward are integrated somewhere in the CNS.

Potential neural substrates where these adjustments may be organized will be

discussed next.

7.2 Central control of posture and movement: integration of

feedback and feedforward postural commands

A primary objective of this thesis was to characterize the organization of

feedforward postural adjustments in order to draw comparisons to the strategies

observed for reactive postural control. The findings of Chapters 4 and 5 present

compelling evidence that the strategies for feedback and feedforward postural

control are similar in their organization. Observation of muscle tuning in pPAs

and aPAs and the existence of muscle synergies for the pPA suggest that an

integration of feedback and feedforward postural commands within the CNS.

Moreover, it is possible that the same neural pathways or motor modules are

recruited for controlling balance in voluntary and reactive postural tasks.

The localization within the CNS of these motor modules, or muscle

synergies, for either feedback or feedforward postural control is not currently

known. Substantial neurophysiological evidence points to the spinal cord (Bizzi

et al. 1991; Hart and Giszter 2004) and brainstem (Deliagina et al. 2008;

Honeycutt et al. 2009; Macpherson et al. 1997) as potential candidates for

encoding motor modules across a wide range of behaviours. For postural

control, however, it appears that the involvement of the brainstem is required.

Supraspinal structures are known to mediate postural responses organized in

both feedback and feedforward (Massion 1992). Furthermore, the lack of

spatial tuning of muscle activity in many postural muscles in spinalized cats

174

subjected to multidirectional support surface perturbations (Macpherson et al.

2007; Macpherson and Fung 1999) suggests that higher levels of the CNS are

required for appropriately mediating the postural muscle activity. Also, lesion

and inactivation studies of the reticulospinal circuitry of the brainstem result in

significant deficits of balance (Gorska et al. 1990; Gorska et al. 1995; Lawrence

and Kuypers 1968; Luccarini et al. 1990).

Recently, neurophysiological studies exploring the nature of the signals

of the brainstem during goal-directed reaching and postural responses to

unexpected disturbances of balance in the cat have shown that signals in the

PMRF contribute to both feedback and feedforward postural control (Schepens

and Drew 2006; 2004; 2003; Schepens et al. 2008; Stapley and Drew 2009).

These findings suggest that the PMRF of the brainstem gates neural signals for

both feedback and feedforward postural control and is the site of integration of

these two signals (Schepens et al. 2008; Stapley and Drew 2009). If so,

similarities in the outward expression of feedback and feedforward postural

behaviour, in terms of patterns of EMG and resulting force, would support this

notion. The brainstem has been proposed as an ideal structure for performing

this integration, and as it receives input from numerous descending commands

and rich afferent feedback from the spinal cord and periphery (Drew et al.

1996). Our results support this hypothesis by providing biomechanical and

electromyographic evidence that posture is organized similarly in both feedback

and feedforward modes of control. We suggest that the muscle synergies for

postural control may be organized at the level of the brainstem.

7.3 Predictive motor control: internal model of posture

A second major aim of the thesis was to examine how posture and

movement components of motor behaviour interact during online corrections of

reaching movements executed in standing. Classic double step paradigms, in

which the light shift is related to a gaze shift, have been used to examine the

visuomotor processes involved in making rapid online corrections to goal-

directed movements (Day and Brown 2001; Day and Lyon 2000; Pélisson et al.

175

1986; Prablanc and Martin 1992). These studies have shown that vision of the

arm improves accuracy by providing a continuous signal of the reach trajectory

in relation to the target goal. Overall, these studies have suggested that the CNS

uses a predictive forward model to rapidly detect deviations from the desired

trajectory and signal a correction (Davidson and Wolpert 2005; Desmurget and

Grafton 2000; Gritsenko et al. 2009). How equilibrium constraints affect the

ability to make such online corrections of reaching movement has received little

attention. Specifically, is it not known if posture is updated in a similar

predictive fashion when online visuomotor perturbations of the reaching

movement are imposed when the subject is standing.

Accordingly, in Chapter 6, a modified double step paradigm was

employed to challenge both the postural and movement systems and examine

the interaction between the two. Although previous studies have shown that

reach uncertainty is integrated in a predictive manner (Martin et al. 2000) and

that corrections to both posture and movement occur automatically (Fautrelle et

al. 2010), neither study systematically addressed the temporal relationships

between the postural and reach behaviours. As such, the analyses performed in

Chapter 6 investigated the temporal relationships between the postural and

focal commands during online corrections of reaching during standing. It was

found that subjects consistently updated their postural commands prior to

correcting their arm movement, even when perturbations were initiated late in

the reaching movement (i.e., deceleration phase of the reach). These results are

interpreted to suggest that the CNS adopts a predictive model of the body and

arm dynamics for controlling posture rather than relying on delayed feedback of

the arm correction. These findings support previous observations that the CNS

is able to modify weight shift during APAs prior to step initiation (Ruget et al.

2008).

The results presented here contribute to a general understanding of the

mechanisms underlying postural control during the execution of voluntary. A

widely accepted view is that the CNS plans goal-directed movements using

knowledge of the dynamics of the musculoskeletal system and the environment

176

(Davidson and Wolpert 2005). Posture is thought to be controlled in a similar

way (Bouisset and Zattara 1981; 1987; Ruget et al. 2008). However, prior to the

study herein, no data could support or refute the hypothesis that the CNS

continued to rely on a predictive internal model during online corrections of the

focal movement. Specifically, the fact that posture was updated prior to the

reach correction would suggest that corrections to posture and movement are

not based upon a purely feedback-driven model of limb movement, as the

postural adjustments occurred in advance of the adjustments to limb trajectory

and muscle activation. Therefore, our results support the notion of a predictive

internal model of voluntary movement, whereby the consequences of the action

are estimated in advance and planned for by the CNS.

7.4 Justification for understanding disorders of posture and

balance

Currently, it is estimated that over one million Canadians experience at

least one fall per year, with associated health care costs estimated at $2.8 billion

annually (Health Canada, 2002). More important, however, are the devastating

consequences of fall-related injuries, which are known to precipitate more

serious, life-threatening co-morbidities. The study of instability and falls has

focused mainly on the mechanisms of reactive postural control in populations

known to exhibit deficits in balance control (Mansfield and Maki 2009; Tokuno

et al. 2010). However, falls commonly occur in situations of dynamic balance,

where both postural and focal components of a movement must be coordinated.

It is not known if individuals fall as a result of deficits in their feedback or

feedforward control mechanisms, or whether they lack the ability to coordinate

posture and movement (Woollacott 2006).

While the emphasis of the presented work has been to understand

feedforward postural control mechanisms associated with reaching movements

in healthy young individuals, the data generated from these studies provide a

basis for characterizing differences in feedforward postural control strategies in

177

people prone to falls. Notably, both the spatial and temporal organization of the

pPAs and aPAs identified in these studies can serve as a benchmark for

comparison with populations afflicted with deficits in balance control.

Furthermore, using the experimental paradigm used here, the online control of

posture during visuomotor perturbations of reaching can be investigated in

different populations. Specifically, the hypothesis that an inability to update

postural commands using a predictive signal when the goal of the focal

movement changes results in a loss of balance could be tested. Thus, both the

experimental paradigm and the results of this thesis have significant

implications for clinical and fundamental applications.

7.5 Conclusions and future directions

Together, the results from this thesis contribute significantly to our

understanding of the central organization of posture and movement.

Furthermore, the experimental paradigm developed for the studies of this thesis

provides a basis for exploring the coordination of posture and movement in

populations suffering from balance deficits. I have shown similarities in the

organization of feedforward postural adjustments to those observed in feedback

postural control, providing biomechanical evidence that the CNS may rely on

similar neural pathways to control posture during predictable and unexpected

disturbances of balance. Subsequently, I demonstrated that the CNS relies on a

predictive control in the aPA when online corrections in movements are

required. While the results presented in this thesis provide important insight

into our understanding of how the CNS coordinates posture and movement, a

number of questions are raised that require further study. Some of these are

reviewed in the following paragraphs.

7.5.1 Do the elderly differ in the spatial and temporal organization of

feedforward postural control?

178

Clear deficits in postural control in the elderly have been documented

(Woollacott 2006). However, most studies have focused on the differences in

reactive postural strategies compared to healthy controls, with little attention on

feedforward postural control. Therefore, to fully understand why and how

people become destabilized and subsequently fail to recover balance with

increasing age, studies exploring the coordination of posture and movement are

required. Additional studies exploring the organization of predictive postural

control in these populations are required if we are to fully understand why and

how people become destabilized and subsequently fail to recover balance. The

experimental paradigm developed here provides the basis for undertaking such

studies.

The task and methodology developed in this thesis provide a novel

means for examining the coordination of posture and movement in the elderly.

These studies would complement our current knowledge about deficits in

feedback postural control. By it’s very nature, the reaching task executed in

standing allows a clear dissociation between the changes related to either the

postural or focal components of the overall movement strategy. Both the

temporal and spatial organization of muscle activity and GRF patterns can be

quantified in detail, and compared to the baseline measures for healthy young

adults documented in Chapters 4 and 5. This approach will make it possible to

identify differences in control strategies that may contribute to instability and

subsequent falls by making comparisons between healthy and afflicted persons.

Using this knowledge, it is hoped that targeted rehabilitation programs for

improving balance control can be developed and refined.

Moreover, it would be possible to determine whether the elderly are

capable of updating posture in a predictive manner when online changes in the

movement goal are required. It is possible that the elderly rely on a feedback

mode of control in such situations, and may be destabilized by the change in the

focal segment’s trajectory. The experiments performed in Chapter 6 could be

extended to elderly populations to test this hypothesis. This would provide

important insight into the control and coordination of posture in dynamic

179

situations in the elderly, who often become destabilized in these types of

dynamic situations. These methods and findings could then be extended to

other populations suffering from deficits in balance control.

7.5.2 Are feedforward muscle synergies robust and how does their

recruitment relate to task-level goals?

In Chapter 5, it was found that the temporal recruitment of a few muscle

synergies can explain the coordination of muscle activity of the pPAs,

presumed to be controlled strictly in feedforward (Massion 1992). The first

question that arises from this study is related to the robustness of the pPA

muscle synergies: are the same muscle synergies recruited for the pPAs as

aPAs? In feedback postural control, muscle synergy robustness has been

evaluated by examining the composition and activation of muscle in different

postural configurations (Torres-Oviedo et al. 2006; Torres-Oviedo and Ting

2010), sensory contexts (Torres-Oviedo and Ting 2010) and types of postural

reactions (Chvatal et al. 2011). Muscle synergies have also been shown to be

stable across a variety of motor behaviours, although task-specific synergies

may be recruited (Cheung et al. 2005; Chvatal et al. 2011; d'Avella and Bizzi

2005). Therefore, I propose to evaluate muscle synergy robustness by

determining whether the CNS recruits the same muscles synergies in the pPA

and aPA periods. Given that the goal of the postural adjustments differs for the

the pPA and aPA (acceleration and deceleration of the CoM, respectively), I

hypothesize that the CNS will recruit a set of shared synergies in addition to

task-specific muscle synergies in the aPA.

In the event that task-specific muscle synergies are found for the aPAs,

it is possible that these additional synergies explain the additional directions of

GRF produced in the aPA period. To test the hypothesis that stabilization of the

body at the target requires additional muscle synergies related to the GRF, a

functional synergy analysis could be performed. To determine these

relationships, biomechanical parameters, such as GRF direction (Ting and

Macpherson 2005; Torres-Oviedo et al. 2006) and CoM acceleration (Chvatal

180

et al. 2011) would be included in the NNMF algorithm. Also, to increase the

richness and variability of the data set, the reaching experiments could be

performed with a variety of stance configurations, such a wide, narrow, natural,

and single foot. This will permit the examination of the muscle synergies in a

variety of biomechanical contexts.

7.5.3 Online control of posture: effects of direction and time of visual

perturbation

The findings of Chapter 6 demonstrated that the postural adjustments

required for initiating online corrections in arm movements while standing for

target shifts from a central target to the right of the midline are organized

predictively of the arm correction. To make these corrections, postural muscle

activity was programmed to create significant loading under the left foot to

break the counter-clockwise rotation and accelerate the body to the new target

(see Fig 6.4B). In the event of a center-left target shift, however, the body

would already be rotating in the direction of the new target. The breaking and

acceleration of the body observed for center-right corrections would not be

required for successful execution of the task in a center-left perturbation. This

begs the question whether the initial planned postural adjustment to the first

target is robust enough to support the change in focal reach direction?

Specifically, would posture continue to be updated predictively in situations

where the body is already in motion towards the corrected target, or would the

adjustments in posture occur in a reactive mode, relying on afferent feedback

related to the motion of the arm or of the CoM in relation to the BoS? To test

the hypothesis that the CNS continues to rely on a predictive mode of control

irrespective of the direction of the perturbation, the postural activity in relation

to the arm EMG and trajectory for center-left perturbations should be

quantified.

A second issue raised from the study in Chapter 6 is how the latency of

the target shift affects the ability of the CNS to organize the appropriate

predictive postural adjustments. In this study, the majority of the target shifts

181

occurred at or very soon after peak velocity during the deceleration phase of the

reaching movement (see Fig 6.1C). It is possible that target shifts occurring

earlier in the execution of the reaching movement may involve a different mode

of control for organizing the postural adjustment. For example, several studies

of perturbed reaching movements executed in seating have shown that the reach

corrections occur ‘automatically’ when target shifts are presented just before or

immediately after the onset of the first gaze shift (Day and Brown 2001; Day

and Lyon 2000; Pélisson et al. 1986; Prablanc and Martin 1992). It remains to

be seen how posture and movement interact when corrections to the reach are

triggered early in the movement, and in relation to visual information, such as a

gaze shift, rather than the reach onset. Therefore, to confirm that a predictive

mode of postural control is consistently used irrespective of latency of the

target shift, experiments involving target shifts occurring at several latencies

with respect to an initial gaze shift should be performed. The findings of such

studies, when compared to online corrections of reaching in seated, will provide

insight into how equilibrium constraints affect the planning and execution of

online corrections involving posture and movement.

182

183

Chapter 8: References

Abdi H. The Bonferonni and Šidák Corrections for Multiple Comparisons. In: Encyclopedia of

measurement and statistics, edited by Salkind N. Thousand Oaks, CA: Sage, 2007, p. 103-107.

Alstermark B, and Isa T. Circuits for Skilled Reaching and Grasping. Annual review of neuroscience 2012.

Aruin AS, and Latash ML. Directional specificity of postural muscles in feed-forward

postural reactions during fast voluntary arm movements. Experimental brain research 103:

323-332, 1995a.

Aruin AS, and Latash ML. The role of motor action in anticipatory postural adjustments

studied with self-induced and externally triggered perturbations. Experimental brain research

106: 291-300, 1995b.

Babinski J. De l'asynergie cerebelleuse. Rev Neurol 7: 806-816, 1899.

Balasubramaniam R, and Turvey M. The λ-Based Equilibrium Point (EP) Hypothesis 1990-

1999. 1-6, 2005.

Bastian AJ. Learning to predict the future: the cerebellum adapts feedforward movement

control. Current Opinion in Neurobiology 16: 645-649, 2006.

Becker W, and Jürgens R. An analysis of the saccadic system by means of double step

stimuli. VISION RESEARCH 19: 967-983, 1979.

Bernstein N. The Coordination and Regulation of Movements. Oxford: Pergamon Press, 1967,

p. 196.

Bhushan N, and Shadmehr R. Computational nature of human adaptive control during

learning of reaching movements in force fields. Biological Cybernetics 81: 39-60, 1999.

Biguer B, Jeannerod M, and Prablanc C. The coordination of eye, head, and arm movements

during reaching at a single visual target. Experimental brain research 46: 301-304, 1982.

Bizzi E, Mussa-Ivaldi FA, and Giszter S. Computations Underlying the Execution of

Movement - A Biological Perspective. Science 253: 287-291, 1991.

Bizzi E, Tresch MC, Saltiel P, and d&apos;Avella A. New perspectives on spinal motor systems. Nature Reviews Neuroscience 1: 101-108, 2000.

Black FO, Shupert CL, Peterka RJ, and Nashner LM. Effects of unilateral loss of vestibular

function on the vestibulo-ocular reflex and postural control. The Annals of otology, rhinology,

and laryngology 98: 884-889, 1989.

Bonnetblanc F, Martin O, and Teasdale N. Pointing to a target from an upright standing

position: anticipatory postural adjustments are modulated by the size of the target in humans.

Neuroscience letters 358: 181-184, 2004.

184

Bouisset S, and Zattara M. A sequence of postural movements precedes voluntary movement.

Neuroscience letters 22: 263-270, 1981.

Bouisset S, and Zattara M. Biomechanical study of the programming of anticipatory postural

adjustments associated with voluntary movement. Journal of Biomechanics 20: 735-742, 1987.

Brustein E, and Rossignol S. Recovery of locomotion after ventral and ventrolateral spinal

lesions in the cat. I. Deficits and adaptive mechanisms. J Neurophysiol 80: 1245-1267, 1998.

Buchanan TS, Almdale DP, Lewis JL, and Rymer WZ. Characteristics of synergic relations

during isometric contractions of human elbow muscles. Journal of neurophysiology 56: 1225-

1241, 1986.

Buchanan TS, Rovai GP, and Rymer WZ. Strategies for muscle activation during isometric

torque generation at the human elbow. Journal of neurophysiology 62: 1201-1212, 1989.

Bunderson NE, Mckay JL, Ting LH, and Burkholder TJ. Directional constraint of endpoint

force emerges from hindlimb anatomy. The Journal of experimental biology 213: 2131-2141,

2010.

Canada H. Healthy Aging: Prevention of Unintentional Injuries Among Seniors. 2002.

Cappellini G, Ivanenko YP, Poppele RE, and Lacquaniti F. Motor patterns in human

walking and running. Journal of neurophysiology 95: 3426-3437, 2006.

Carpenter MG, Allum JH, and Honegger F. Directional sensitivity of stretch reflexes and balance corrections for normal subjects in the roll and pitch planes. Experimental brain

research 129: 93-113, 1999.

Cavagna GA, Heglund NC, and Taylor CR. Mechanical work in terrestrial locomotion: two

basic mechanisms for minimizing energy expenditure. The American journal of physiology 233:

R243-261, 1977.

Cavanagh P, Derr J, Ulbrecht J, Maser R, and Orchard T. Problems with gait and posture

in neuropathic patients with insulin-dependent diabetes mellitus. Diabet Med 9: 469-474, 1992.

Chanaud CM, and Macpherson JM. Functionally complex muscles of the cat hindlimb. 3.

Differential activation within biceps femoris during postural perturbations. Exp Brain Res 85:

271-280, 1991.

Cheung VCK, d'Avella A, Tresch MC, and Bizzi E. Central and sensory contributions to the activation and organization of muscle synergies during natural motor behaviors. Journal of

Neuroscience 25: 6419-6434, 2005.

Chiel HJ, Ting LH, Ekeberg O, and Hartmann MJZ. The brain in its body: motor control

and sensing in a biomechanical context. Journal of Neuroscience 29: 12807-12814, 2009.

Chvatal SA, Torres-Oviedo G, Safavynia SA, and Ting LH. Common muscle synergies for

control of center of mass and force in nonstepping and stepping postural behaviors. Journal of

neurophysiology 106: 999-1015, 2011.

Commissaris DA, Toussaint HM, and Hirschfeld H. Anticipatory postural adjustments in a

bimanual, whole-body lifting task seem not only aimed at minimising anterior--posterior centre

of mass displacements. Gait & Posture 14: 44-55, 2001.

185

Cordo PJ, and Nashner LM. Properties of postural adjustments associated with rapid arm

movements. Journal of neurophysiology 47: 287-302, 1982.

d'Avella A, and Bizzi E. Shared and specific muscle synergies in natural motor behaviors.

Proceedings of the National Academy of Sciences of the United States of America 102: 3076-

3081, 2005.

d'Avella A, Portone A, and Lacquaniti F. Superposition and modulation of muscle synergies for reaching in response to a change in target location. Journal of neurophysiology 106: 2796-

2812, 2011.

Danion F, and Sarlegna FR. Can the human brain predict the consequences of arm movement

corrections when transporting an object? Hints from grip force adjustments. Journal of

Neuroscience 27: 12839-12843, 2007.

Davidson PR, and Wolpert DM. Widespread access to predictive models in the motor system:

a short review. Journal of neural engineering 2: S313-319, 2005.

Day BI, and Brown P. Evidence for subcortical involvement in the visual control of human

reaching. Brain 124: 1832-1840, 2001.

Day BL, and Lyon IN. Voluntary modification of automatic arm movements evoked by

motion of a visual target. Experimental brain research 130: 159-168, 2000.

Day BL, Severac Cauquil A, Bartolomei L, Pastor MA, and Lyon IN. Human body-segment tilts induced by galvanic stimulation: a vestibularly driven balance protection mechanism. J

Physiol 500 ( Pt 3): 661-672, 1997.

Deliagina TG, Beloozerova IN, Zelenin PV, and Orlovsky GN. Spinal and supraspinal

postural networks. Brain research reviews 57: 212-221, 2008.

Desmurget M, and Grafton S. Forward modeling allows feedback control for fast reaching

movements. Trends in cognitive sciences 4: 423-431, 2000.

Diener HC, Dichgans J, Bootz F, and Bacher M. Early stabilization of human posture after a

sudden disturbance: influence of rate and amplitude of displacement. Experimental brain

research 56: 126-134, 1984a.

Diener HC, Dichgans J, Guschlbauer B, and Mau H. The significance of proprioception on

postural stabilization as assessed by ischemia. Brain research 296: 103-109, 1984b.

Drew T, Cabana T, and Rossignol S. Responses of medullary reticulospinal neurones to

stimulation of cutaneous limb nerves during locomotion in intact cats. Experimental Brain

Research 111: 153-168, 1996.

Drew T, Jiang W, and Widajewicz W. Contributions of the motor cortex to the control of the

hindlimbs during locomotion in the cat. Brain research Brain research reviews 40: 178-191,

2002.

Drew T, Kalaska J, and Krouchev N. Muscle synergies during locomotion in the cat: a model

for motor cortex control. The Journal of Physiology 586: 1239-1245, 2008.

Drew T, Prentice S, and Schepens B. Cortical and brainstem control of locomotion. Progress

in brain research 143: 251-261, 2004.

186

Duff SV, and Sainburg RL. Lateralization of motor adaptation reveals independence in

control of trajectory and steady-state position. Experimental brain research 179: 551-561,

2007.

Eccles JC, and Sherrington CS. Reflex summation in the ipsilateral spinal flexion reflex. The

Journal of Physiology 69: 1-28, 1930.

Elliott D, Garson RG, Goodman D, and Chua R. Discrete vs. continuous visual control of manual aiming. Human Movement Science 10: 393-418, 1991.

Fautrelle L, Prablanc C, Berret B, Ballay Y, and Bonnetblanc F. Pointing to double-step

visual stimuli from a standing position: very short latency (express) corrections are observed in

upper and lower limbs and may not require cortical involvement. Neuroscience 169: 697-705,

2010.

Flanders M, Daghestani L, and Berthoz A. Reaching beyond reach. Experimental brain

research 126: 19-30, 1999.

Flanders M, and Soechting JF. Arm muscle activation for static forces in three-dimensional

space. Journal of neurophysiology 64: 1818-1837, 1990.

Gahery Y, and Nieoullon A. Postural and kinetic coordination following cortical stimuli which

induce flexion movements in the cat&apos;s limbs. Brain research 149: 25-37, 1978.

Georgopoulos AP. On reaching. Annual review of neuroscience 9: 147-170, 1986.

Georgopoulos AP, Kalaska JF, and Massey JT. Spatial trajectories and reaction times of

aimed movements: effects of practice, uncertainty, and change in target location. J

Neurophysiol 46: 725-743, 1981.

Ghez C, and Krakauer JW. The organization of movement. In: Principles of neural science,

edited by Kandel ER, Schwartz JH, and Jessell TM. New York: McGraw-Hill, 2000, p. 653-

673.

Gielen CC, van den Heuvel PJ, and van Gisbergen JA. Coordination of fast eye and arm

movements in a tracking task. Experimental brain research 56: 154-161, 1984.

Goodale MA, Pelisson D, and Prablanc C. Large adjustments in visually guided reaching do

not depend on vision of the hand or perception of target displacement. Nature 320: 748-750,

1986.

Gorska T, Bem T, and Majczynski H. Locomotion in Cats with Ventral Spinal Lesions -

Support Patterns and Duration of Support Phases During Unrestrained Walking. Acta Neurobiol

Exp 50: 191-200, 1990.

Gorska T, Ioffe M, Zmyslowski W, Bem T, Majczynski H, and Mats VN. Unrestrained

walking in cats with medial pontine reticular lesions. Brain Research Bulletin 38: 297-304,

1995.

Grillner S. Biological pattern generation: the cellular and computational logic of networks in

motion. Neuron 52: 751-766, 2006.

Grillner S. The motor infrastructure: from ion channels to neuronal networks. Nature Reviews

Neuroscience 4: 573-586, 2003.

187

Gritsenko V, Yakovenko S, and Kalaska JF. Integration of predictive feedforward and

sensory feedback signals for online control of visually guided movement. Journal of

neurophysiology 102: 914-930, 2009.

Gutierrez-Farewik EM, Bartonek A, and Saraste H. Comparison and evaluation of two

common methods to measure center of mass displacement in three dimensions during gait.

Human Movement Science 25: 238-256, 2006.

Hart CB, and Giszter SF. Modular premotor drives and unit bursts as primitives for frog

motor behaviors. Journal of Neuroscience 24: 5269-5282, 2004.

Hayes KC. Biomechanics of postural control. Exercise and sport sciences reviews 10: 363-391,

1982.

Henry SM, Fung J, and Horak FB. Control of stance during lateral and anterior/posterior

surface translations. IEEE transactions on rehabilitation engineering : a publication of the

IEEE Engineering in Medicine and Biology Society 6: 32-42, 1998a.

Henry SM, Fung J, and Horak FB. Effect of stance width on multidirectional postural

responses. Journal of neurophysiology 85: 559-570, 2001.

Henry SM, Fung J, and Horak FB. EMG responses to maintain stance during multidirectional

surface translations. Journal of neurophysiology 80: 1939-1950, 1998b.

Herrmann U, and Flanders M. Directional tuning of single motor units. The Journal of neuroscience : the official journal of the Society for Neuroscience 18: 8402-8416, 1998.

Hess W. Telekinetisches und ereismatisches r ftesystem in der Biomotorik. . Helv Physiol

Pharmacol Acta 1: C62-C63, 1954.

Hess W. Teleokinetisches und ereismatisches Kräftesystem in der Biomotorik? Helv Physiol

Pharmac Acta 1: C62-C63, 1943.

Hollands MA, Ziavra NV, and Bronstein AM. A new paradigm to investigate the roles of

head and eye movements in the coordination of whole-body movements. Experimental brain

research 154: 261-266, 2004.

Hollerbach MJ, and Flash T. Dynamic interactions between limb segments during planar arm

movement. Biological Cybernetics 44: 67-77, 1982.

Holmquist B, and Sandberg R. The broken axis approach - a new way to analyze bi-directional circular data. Experientia 47: 845-851, 1991.

Honeycutt CF, Gottschall JS, and Nichols TR. Electromyographic responses from the

hindlimb muscles of the decerebrate cat to horizontal support surface perturbations. Journal of

neurophysiology 101: 2751-2761, 2009.

Honeycutt CF, and Nichols TR. The decerebrate cat generates the essential features of the

force constraint strategy. Journal of neurophysiology 103: 3266-3273, 2010.

Horak FB, Diener HC, and Nashner LM. Influence of central set on human postural

responses. Journal of neurophysiology 62: 841-853, 1989a.

188

Horak FB, and Macpherson JM. Postural orientation and equilibrium. In: Handbook of

Physiology Section 12: Exercise: Regulation and Integration of Multiple Systems, edited by

Rowell LB, and Shepherd JT. New York, NY: Oxford Univ Press, 1996, p. 255-292.

Horak FB, and Moore SP. The effect of prior leaning on human postural responses. Gait &

Posture 1: 203-210, 1993.

Horak FB, and Nashner LM. Central programming of postural movements: adaptation to altered support-surface configurations. Journal of neurophysiology 55: 1369-1381, 1986.

Horak FB, Nashner LM, and Diener HC. Postural strategies associated with somatosensory

and vestibular loss. Experimental brain research 82: 167-177, 1990.

Horak FB, Shupert CL, and Mirka A. Components of postural dyscontrol in the elderly: a

review. Neurobiology of aging 10: 727-738, 1989b.

Inglis JT, and Macpherson JM. Bilateral labyrinthectomy in the cat: effects on the postural

response to translation. Journal of neurophysiology 73: 1181-1191, 1995.

Inglis JT, and Macpherson JM. The role of the cat vestibular system in the response to linear

acceleration of the support surface. Neurosci Abstr 14: 63, 1988.

Ivanenko YP, Cappellini G, Dominici N, Poppele RE, and Lacquaniti F. Coordination of

locomotion with voluntary movements in humans. Journal of Neuroscience 25: 7238-7253,

2005.

Johansson RS, and Westling G. Roles of glabrous skin receptors and sensorimotor memory in

automatic control of precision grip when lifting rougher or more slippery objects. Exp Brain

Res 56: 550-564, 1984.

Joseph J, and Nightingale A. Electromyography of muscles of posture: Leg muscles in males.

J Physiol 117: 484-491, 1952.

Kably B, and Drew T. Corticoreticular pathways in the cat. I. Projection patterns and

collaterization. Journal of neurophysiology 80: 389-405, 1998.

Kawato M. Internal models for motor control and trajectory planning. Current Opinion in

Neurobiology 9: 718-727, 1999.

Klous M, Mikulic P, and Latash ML. Early postural adjustments in preparation to whole-

body voluntary sway. Journal of electromyography and kinesiology : official journal of the International Society of Electrophysiological Kinesiology 22: 110-116, 2012.

Komilis E, Pélisson D, and Prablanc C. Error processing in pointing at randomly feedback-

induced double-step stimuli. Journal of motor behavior 25: 299-308, 1993.

Krishnamoorthy V, Goodman S, Zatsiorsky V, and Latash ML. Muscle synergies during

shifts of the center of pressure by standing persons: identification of muscle modes. Biological

Cybernetics 89: 152-161, 2003.

Krishnamoorthy V, Latash ML, Scholz JP, and Zatsiorsky VM. Muscle modes during shifts

of the center of pressure by standing persons: effect of instability and additional support.

Experimental brain research Experimentelle Hirnforschung Expérimentation cérébrale 157:

18-31, 2004.

189

Kuo AD. An optimal state estimation model of sensory integration in human postural balance.

Journal of neural engineering 2: S235-249, 2005.

Latash ML. The bliss (not the problem) of motor abundance (not redundancy). Experimental

brain research Experimentelle Hirnforschung Expérimentation cérébrale 2012.

Lawrence DG, and Kuypers HGJM. The functional organization of the motor system in the

monkey. II. The effects of lesions of the descending brain-stem pathways. Brain 91: 15-36, 1968.

Lee DD, and Seung HS. Learning the parts of objects by non-negative matrix factorization.

Nature 401: 788-791, 1999.

Lee WA, Buchanan TS, and Rogers MW. Effects of arm acceleration and behavioral

conditions on the organization of postural adjustments during arm flexion. Experimental brain

research Experimentelle Hirnforschung Expérimentation cérébrale 66: 257-270, 1987.

Leonard JA, Brown RH, and Stapley PJ. Reaching to multiple targets when standing: the

spatial organization of feedforward postural adjustments. Journal of neurophysiology 101:

2120-2133, 2009.

Leonard JA, Gritsenko V, Ouckama R, and Stapley PJ. Postural adjustments for online

corrections of arm movements in standing humans. Journal of neurophysiology 105: 2375-

2388, 2011.

Lepers R, and Brenière Y. The role of anticipatory postural adjustments and gravity in gait

initiation. Experimental brain research Experimentelle Hirnforschung Expérimentation

cérébrale 107: 118-124, 1995.

Levine AJ, Lewallen KA, and Pfaff SL. Spatial organization of cortical and spinal neurons

controlling motor behavior. Current Opinion in Neurobiology 2012.

Liddell EGT, and Sherrington C. Reflexes in Response to Stretch (Myotatic Reflexes).

Proceedings of the Royal Society B: Biological Sciences 96: 212-242, 1924.

Lockhart DB, and Ting LH. Optimal sensorimotor transformations for balance. Nature

Neuroscience 10: 1329-1336, 2007.

Luccarini P, Gahery Y, and Pompeiano O. Cholinoceptive pontine reticular structures

modify the postural adjustments during the limb movement induced by cortical stimulation. Arch Ital Biol 128: 19-45, 1990.

Mackinnon CD, Bissig D, Chiusano J, Miller E, Rudnick L, Jager C, Zhang Y, Mille M-L,

and Rogers MW. Preparation of anticipatory postural adjustments prior to stepping. Journal of

neurophysiology 97: 4368-4379, 2007.

Macpherson J. How flexible are muscle synergies? In: Motor control: Concepts and Issues,

edited by Humphrey DR, and Freund H-J. UK: John Wiley & Sons, 1991, p. 33-47.

Macpherson JM. Strategies that simplify the control of quadrupedal stance. I. Forces at the

ground. Journal of neurophysiology 60: 204-217, 1988a.

Macpherson JM. Strategies that simplify the control of quadrupedal stance. II.

Electromyographic activity. Journal of neurophysiology 60: 218-231, 1988b.

190

Macpherson JM. The force constraint strategy for stance is independent of prior experience.

Experimental brain research Experimentelle Hirnforschung Expérimentation cérébrale 101:

397-405, 1994.

Macpherson JM, Everaert DG, Stapley PJ, and Ting LH. Bilateral Vestibular Loss in Cats

Leads to Active Destabilization of Balance During Pitch and Roll Rotations of the Support

Surface. Journal of neurophysiology 97: 4357-4367, 2007.

Macpherson JM, and Fung J. Weight support and balance during perturbed stance in the

chronic spinal cat. Journal of neurophysiology 82: 3066-3081, 1999.

Macpherson JM, Fung J, and Jacobs R. Postural orientation, equilibrium and the spinal cord.

In: Neuronal Regeneration Reorganization and Repair, edited by Seil F. Philadelphia:

Lippincott-Raven, 1997, p. 227-232.

Maki BE, and McIlroy WE. The role of limb movements in maintaining upright stance: the

"change-in-support" strategy. Physical Therapy 77: 488-507, 1997.

Mansfield A, and Maki BE. Are age-related impairments in change-in-support balance

reactions dependent on the method of balance perturbation? Journal of Biomechanics 42: 1023-

1031, 2009.

Martin O, Teasdale N, Simoneau M, Corbeil P, and Bourdin C. Pointing to a target from an

upright position in human: tuning of postural responses when there is target uncertainty. Neuroscience letters 281: 53-56, 2000.

Massion J. Movement, posture and equilibrium: Interaction and coordination. Progress in

neurobiology 38: 35-56, 1992.

Massion J. Postural control systems in developmental perspective. Neuroscience and

biobehavioral reviews 22: 465-472, 1998.

Massion J, Alexandrov A, and Frolov A. Why and how are posture and movement

coordinated? Progress in brain research 143: 13-27, 2004.

Massion J, and Woollacott MH. Posture and equilibrium. 2004, p. 1-19.

Maurer C, Mergner T, Bolha B, and Hlavacka F. Human balance control during cutaneous

stimulation of the plantar soles. Neuroscience letters 302: 45-48, 2001.

McKay JL, Burkholder TJ, and Ting LH. Biomechanical capabilities influence postural control strategies in the cat hindlimb. Journal of Biomechanics 40: 2254-2260, 2007.

McKay JL, and Ting LH. Functional muscle synergies constrain force production during

postural tasks. Journal of Biomechanics 41: 299-306, 2008.

McVea DA, and Pearson KG. Object avoidance during locomotion. Advances in experimental

medicine and biology 629: 293-315, 2009.

Megaw ED. Possible modification to a rapid on-going programmed manual response. Brain

research 71: 425-441, 1974.

Mehta B, and Schaal S. Forward Models in Visuomotor Control. Journal of neurophysiology

88: 942-953, 2002.

191

Mergner T, and Rosemeier T. Interaction of vestibular, somatosensory and visual signals for

postural control and motion perception under terrestrial and microgravity conditions--a

conceptual model. Brain research Brain research reviews 28: 118-135, 1998.

Milner TE. A model for the generation of movements requiring endpoint precision.

Neuroscience 49: 487-496, 1992.

Mink JW. The basal ganglia: focused selection and inhibition of competing motor programs. Progress in neurobiology 50: 381-425, 1996.

Moore SP, Rushmer DS, Windus SL, and Nashner LM. Human automatic postural

responses: responses to horizontal perturbations of stance in multiple directions. Experimental

brain research Experimentelle Hirnforschung Expérimentation cérébrale 73: 648-658, 1988.

Morasso PG, and Schieppati M. Can muscle stiffness alone stabilize upright standing? J

Neurophysiol 82: 1622-1626, 1999.

Mouchnino L, Aurenty R, Massion J, and Pedotti A. Coordination between equilibrium and

head-trunk orientation during leg movement: a new strategy build up by training. Journal of

neurophysiology 67: 1587-1598, 1992.

Muceli S, Boye AT, d'Avella A, and Farina D. Identifying representative synergy matrixes

for describing muscular activation patterns during multi-directional reaching in the horizontal

plane. Journal of neurophysiology 2010.

Nardone A, and Schieppati M. Postural adjustments associated with voluntary contraction of

leg muscles in standing man. Experimental brain research Experimentelle Hirnforschung

Expérimentation cérébrale 69: 469-480, 1988.

Nashner LM. Fixed patterns of rapid postural responses among leg muscles during stance.

Experimental brain research Experimentelle Hirnforschung Expérimentation cérébrale 30: 13-

24, 1977.

Nashner LM, Black FO, and Wall C. Adaptation to altered support and visual conditions

during stance: patients with vestibular deficits. The Journal of neuroscience : the official

journal of the Society for Neuroscience 2: 536-544, 1982.

Nijhof E. On-line trajectory modifications of planar, goal-directed arm movements. Human

Movement Science 22: 13-36, 2003.

Paulignan Y, Dufossé M, Hugon M, and Massion J. Acquisition of co-ordination between

posture and movement in a bimanual task. Experimental brain research Experimentelle

Hirnforschung Expérimentation cérébrale 77: 337-348, 1989.

Paulignan Y, MacKenzie C, Marteniuk R, and Jeannerod M. The coupling of arm and

finger movements during prehension. Experimental brain research Experimentelle

Hirnforschung Expérimentation cérébrale 79: 431-435, 1990.

Pearson KG, and Gordon J. Spinal reflexes. In: Principles of Neural Science, edited by

Kandel ER, Schwartz JH, and Jessell TM. New York: McGraw Hill, 2000, p. 713-736.

Pélisson D, Prablanc C, Goodale MA, and Jeannerod M. Visual control of reaching

movements without vision of the limb. II. Evidence of fast unconscious processes correcting the

trajectory of the hand to the final position of a double-step stimulus. Experimental brain

research Experimentelle Hirnforschung Expérimentation cérébrale 62: 303-311, 1986.

192

Perfiliev S. Bilateral processing of motor commands in the motor cortex of the cat during

target-reaching. Journal of neurophysiology 93: 2489-2506, 2005.

Peterka RJ. Sensorimotor integration in human postural control. Journal of neurophysiology

88: 1097-1118, 2002.

Prablanc C, and Martin O. Automatic control during hand reaching at undetected two-

dimensional target displacements. Journal of neurophysiology 67: 455-469, 1992.

Prablanc C, Pélisson D, and Goodale MA. Visual control of reaching movements without

vision of the limb. I. Role of retinal feedback of target position in guiding the hand.

Experimental brain research Experimentelle Hirnforschung Expérimentation cérébrale 62:

293-302, 1986.

Purves D, Augustine GJ, Fitzpatrick D, Hall WC, Lamantia A-S, and White LE.

Neuroscience. Sinauer Associates, 2012, p. 858.

Reynolds R, and Day B. Rapid visuo-motor processes drive the leg regardless of balance

constraints. Current biology : CB 15: R48-49, 2005a.

Reynolds R, and Day B. Visual guidance of the human foot during a step. The Journal of

Physiology 569: 677-684, 2005b.

Rogers MW, and Pai YC. Dynamic transitions in stance support accompanying leg flexion

movements in man. Experimental brain research Experimentelle Hirnforschung Expérimentation cérébrale 81: 398-402, 1990.

Roh J, Cheung VCK, and Bizzi E. Modules in the brain stem and spinal cord underlying

motor behaviors. Journal of neurophysiology 106: 1363-1378, 2011.

Rossetti Y, Desmurget M, and Prablanc C. Vectorial coding of movement: vision,

proprioception, or both? Journal of neurophysiology 74: 457-463, 1995.

Rossignol S, Bouyer L, Barthélemy D, Langlet C, and Leblond H. Recovery of locomotion

in the cat following spinal cord lesions. Brain research Brain research reviews 40: 257-266,

2002.

Rossignol S, Dubuc R, and Gossard J-P. Dynamic sensorimotor interactions in locomotion.

Physiological reviews 86: 89-154, 2006.

Ruget H, Blouin J, Teasdale N, and Mouchnino L. Can prepared anticipatory postural adjustments be updated by proprioception? Neuroscience 155: 640-648, 2008.

Sabes PN. The planning and control of reaching movements. Current Opinion in Neurobiology

10: 740-746, 2000.

Safavynia SA, and Ting LH. Task-level feedback can explain temporal recruitment of

spatially-fixed muscle synergies throughout postural perturbations. Journal of neurophysiology

2011.

Safavynia SA, Torres-Oviedo G, and Ting LH. Muscle Synergies: Implications for Clinical

Evaluation and Rehabilitation of Movement. Topics in spinal cord injury rehabilitation 17: 16-

24, 2011.

Saltzman E. Levels of sensorimotor representation. J Math Psychol 20: 91-163, 1979.

193

Santos MJ, and Aruin AS. Role of lateral muscles and body orientation in feedforward

postural control. Experimental brain research Experimentelle Hirnforschung Expérimentation

cérébrale 184: 547-559, 2008.

Sarlegna FR, and Sainburg RL. The roles of vision and proprioception in the planning of

reaching movements. Advances in experimental medicine and biology 629: 317-335, 2009.

Saunders J, and Knill D. Humans use continuous visual feedback from the hand to control fast reaching movements. Experimental brain research Experimentelle Hirnforschung

Expérimentation cérébrale 152: 341-352, 2003.

Schepens B, and Drew T. Descending signals from the pontomedullary reticular formation are

bilateral, asymmetric, and gated during reaching movements in the cat. Journal of

neurophysiology 96: 2229-2252, 2006.

Schepens B, and Drew T. Independent and convergent signals from the pontomedullary

reticular formation contribute to the control of posture and movement during reaching in the

cat. Journal of neurophysiology 92: 2217-2238, 2004.

Schepens B, and Drew T. Strategies for the integration of posture and movement during

reaching in the cat. Journal of neurophysiology 90: 3066-3086, 2003.

Schepens B, Stapley P, and Drew T. Neurons in the pontomedullary reticular formation signal

posture and movement both as an integrated behavior and independently. Journal of neurophysiology 100: 2235-2253, 2008.

Scott SH, Sergio LE, and Kalaska JF. Reaching movements with similar hand paths but

different arm orientations. II. Activity of individual cells in dorsal premotor cortex and parietal

area 5. Journal of neurophysiology 78: 2413-2426, 1997.

Sergio LE, and Kalaska JF. Changes in the temporal pattern of primary motor cortex activity

in a directional isometric force versus limb movement task. Journal of neurophysiology 80:

1577-1583, 1998.

Shabbott BA, and Sainburg RL. On-line corrections for visuomotor errors. Experimental

brain research Experimentelle Hirnforschung Expérimentation cérébrale 195: 59-72, 2009.

Shadmehr R, and Krakauer JW. A computational neuroanatomy for motor control.

Experimental brain research Experimentelle Hirnforschung Expérimentation cérébrale 185: 359-381, 2008.

Sherrington C. The Integrative Action of the Nervous System. New Haven: Yale University

Press, 1961, p. 413.

Sherrington CS. A mammalian spinal preparation. The Journal of Physiology 38: 375-383,

1909.

Shim JK, Olafsdottir H, Zatsiorsky VM, and Latash ML. The emergence and disappearance

of multi-digit synergies during force-production tasks. Experimental brain research 164: 260-

270, 2005.

Soechting JF. Effect of target size on spatial and temporal characteristics of a pointing

movement in man. Experimental brain research Experimentelle Hirnforschung

Expérimentation cérébrale 54: 121-132, 1984.

194

Soechting JF, and Lacquaniti F. Modification of trajectory of a pointing movement in

response to a change in target location. Journal of neurophysiology 49: 548-564, 1983.

Stapley PJ, and Drew T. The pontomedullary reticular formation contributes to the

compensatory postural responses observed following removal of the support surface in the

standing cat. Journal of neurophysiology 101: 1334-1350, 2009.

Stapley PJ, Pozzo T, Cheron G, and Grishin A. Does the coordination between posture and movement during human whole-body reaching ensure center of mass stabilization?

Experimental brain research Experimentelle Hirnforschung Expérimentation cérébrale 129:

134-146, 1999.

Stapley PJ, Ting LH, Hulliger M, and Macpherson JM. Automatic postural responses are

delayed by pyridoxine-induced somatosensory loss. Journal of Neuroscience 22: 5803-5807,

2002.

Thoroughman KA, and Shadmehr R. Electromyographic correlates of learning an internal

model of reaching movements. Journal of Neuroscience 19: 8573-8588, 1999.

Ting LH. Dimensional reduction in sensorimotor systems: a framework for understanding

muscle coordination of posture. Progress in brain research 165: 299-321, 2007.

Ting LH, and Chvatal SA. Decomposing muscle activity in motor tasks. In: Motor Control:

Theories, Experiments, and Applications, edited by Latash ML, and Danion FOxford University Press, 2010, p. 102-140.

Ting LH, Kautz SA, Brown DA, and Zajac FE. Phase reversal of biomechanical functions

and muscle activity in backward pedaling. Journal of neurophysiology 81: 544-551, 1999.

Ting LH, and Macpherson JM. A limited set of muscle synergies for force control during a

postural task. Journal of neurophysiology 93: 609-613, 2005.

Ting LH, and Macpherson JM. Ratio of shear to load ground-reaction force may underlie the

directional tuning of the automatic postural response to rotation and translation. Journal of

neurophysiology 92: 808-823, 2004.

Ting LH, and Mckay JL. Neuromechanics of muscle synergies for posture and movement.

Current Opinion in Neurobiology 17: 622-628, 2007.

Tokuno CD, Cresswell AG, Thorstensson A, and Carpenter MG. Age-related changes in postural responses revealed by support-surface translations with a long acceleration-

deceleration interval. Clinical neurophysiology : official journal of the International Federation

of Clinical Neurophysiology 121: 109-117, 2010.

Torres-Oviedo G, Macpherson JM, and Ting LH. Muscle synergy organization is robust

across a variety of postural perturbations. Journal of neurophysiology 96: 1530-1546, 2006.

Torres-Oviedo G, and Ting LH. Muscle synergies characterizing human postural responses.

Journal of neurophysiology 98: 2144-2156, 2007.

Torres-Oviedo G, and Ting LH. Subject-specific muscle synergies in human balance control

are consistent across different biomechanical contexts. Journal of neurophysiology 103: 3084-

3098, 2010.

195

Tresch MC, Cheung VCK, and d'Avella A. Matrix factorization algorithms for the

identification of muscle synergies: evaluation on simulated and experimental data sets. Journal

of neurophysiology 95: 2199-2212, 2006.

Tresch MC, and Jarc A. The case for and against muscle synergies. Current Opinion in

Neurobiology 19: 601-607, 2009.

Tresch MC, Saltiel P, and Bizzi E. The construction of movement by the spinal cord. Nature Neuroscience 2: 162-167, 1999.

van Ingen Schenau GJ, Boots PJ, de Groot G, Snackers RJ, and van Woensel WW. The

constrained control of force and position in multi-joint movements. Neuroscience 46: 197-207,

1992.

Viallet F, Massion J, Massarino R, and Khalil R. Coordination between posture and

movement in a bimanual load lifting task: putative role of a medial frontal region including the

supplementary motor area. Experimental brain research Experimentelle Hirnforschung

Expérimentation cérébrale 88: 674-684, 1992.

von Holst E, and Mittelstaedt H. Das reafferenzprinzip [Translation of title: The principle of

reafference]. Naturwissenschaften 37: 464-476, 1950.

Welch TDJ, and Ting LH. A feedback model reproduces muscle activity during human

postural responses to support-surface translations. Journal of neurophysiology 99: 1032-1038, 2008.

Westheimer G. Eye movement responses to a horizontally moving visual stimulus. Arch

Ophthalmol 52: 932-941, 1954.

Wheeless LL, Boynton RM, and Cohen GH. Eye-movement responses to step and pulse-step

stimuli. Journal of the Optical Society of America 56: 956-960, 1966.

Winter DA. Biomechanics and Motor Control of Human Movement. John Wiley and Sons,

2009, p. 384.

Winter DA, and Eng P. Kinetics: our window into the goals and strategies of the central

nervous system. Behavioural brain research 67: 111-120, 1995.

Wolpert D, and Miall R. Forward Models for Physiological Motor Control. Neural networks :

the official journal of the International Neural Network Society 9: 1265-1279, 1996.

Woollacott MH. Changes in posture and voluntary control in the elderly: Research findings

and rehabilitation. Topics in Geriatric Rehabilitation 5: 1-11, 2006.

Yakovenko S, and Drew T. A motor cortical contribution to the anticipatory postural

adjustments that precede reaching in the cat. Journal of neurophysiology 102: 853-874, 2009.

Zar JH. Biostatistical Analysis. New Jersey: Prentice Hall, 1999, p. 663.

Zelaznik HZ, Hawkins B, and Kisselburgh L. Rapid visual feedback processing in single-

aiming movements. Journal of motor behavior 15: 217-236, 1983.