Human Movement Science 39 (2015) 121137

Contents lists available at ScienceDirect

Human Movement Science

journal homepage: www.elsevier.com/locate/humov

Human pelvis motions when walking and when

riding a therapeutic horse
Brian A. Garner a,, B. Rhett Rigby b,1
a
b

Department of Mechanical Engineering, Baylor University, Waco, TX, USA

Department of Health, Human Performance, and Recreation, Baylor University, Waco, TX, USA

a r t i c l e

i n f o

Article history:

PsycINFO classication:
2330
2221
Keywords:
Equine assisted activities
Therapeutic horse riding
Pelvis kinematics
Human gait
Motion capture

a b s t r a c t
A prevailing rationale for equine assisted therapies is that the
motion of a horse can provide sensory stimulus and movement
patterns that mimic those of natural human activities such as
walking. The purpose of this study was to quantitatively measure
and compare human pelvis motions when walking to those when
riding a horse. Six able-bodied children (inexperienced riders, 8
12 years old) participated in over-ground trials of self-paced walking and leader-paced riding on four different horses. Five kinematic
measures were extracted from three-dimensional pelvis motion
data: anteroposterior, superoinferior, and mediolateral translations, list angle about the anteroposterior axis, and twist angle
about the superoinferior axis. There was generally as much or more
variability in motion range observed between riding on the different horses as between riding and walking. Pelvis trajectories
exhibited many similar features between walking and riding,
including distorted lemniscate patterns in the transverse and frontal planes. In the sagittal plane the pelvis trajectory during walking
exhibited a somewhat circular pattern whereas during riding it
exhibited a more diagonal pattern. This study shows that riding
on a horse can generate movement patterns in the human pelvis
that emulate many, but not all, characteristics of those during natural walking.
2014 Elsevier B.V. All rights reserved.

Corresponding author at: One Bear Place #97356, Baylor University, Waco, TX 76798-7356, USA. Tel.: +1 254 710 4191; fax:
+1 254 710 3360.
E-mail addresses: Brian_Garner@baylor.edu (B.A. Garner), Rhett_Rigby@baylor.edu (B.R. Rigby).
1
Address: One Bear Place #97313, Baylor University, Waco, TX 76798-7313, USA. Tel.: +1 254 710 3242; fax: +1 254 710 3527.
http://dx.doi.org/10.1016/j.humov.2014.06.011
0167-9457/ 2014 Elsevier B.V. All rights reserved.

122

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

1. Introduction
Motor practice and the repetition of cyclic motions are important for the establishment, development, reinforcement, and improvement of neural and motor pathways (Casady & Nichols-Larsen,
2004; Damiano, 2006; Miller et al., 2010; Willoughby, Dodd, Shields, & Foley, 2010). Therapeutic intervention programs that emphasize repetitive practice of skilled activities or cyclic motions such as
walking may be helpful for many types of neurological disorders including those associated with cerebral palsy, stroke, spinal cord injuries, and other central nervous system disorders (Damiano, 2006;
Ketelaar et al., 2001; Miller et al., 2010; Willoughby et al., 2010). Such therapies may be administered
through manual assistance of a therapist (Geiger, Allen, OKeefe, & Hicks, 2001; Shurtleff, Standeven, &
Engsberg, 2009; Tecklin, 2007) or through assistance with various mechanical devices such as exercise
machines (Dodd, Taylor, & Damiano, 2002; Miller et al., 2010), treadmills (Damiano, 2006; Miller et al.,
2010; Stanger & Oresic, 2003; Willoughby et al., 2010), or even robots (Hogan et al., 2006; Miller et al.,
2010; Veneman et al., 2007).
Equine assisted therapy (EAT) is another therapeutic treatment strategy that emphasizes repetitive
motion. Also called hippotherapy, these treatment programs are designed by a licensed health professional and include time riding on a horse as part of an integrated intervention strategy (American
Hippotherapy Association, 2010). These programs use the cyclic, multidimensional motion of a horse
and the resulting patient responses to help achieve functional outcomes for patients with a wide variety of neuromusculoskeletal dysfunctions including cerebral palsy, sensory integration disorders,
traumatic brain injury, stroke, and others (American Hippotherapy Association, 2010). There is growing, quantitative evidence in the literature supporting the benets of EAT for regulating muscle tone
and improving range-of-motion, spasticity, muscle symmetry, balance, postural control, and other
abilities (Benda, McGibbon, & Grant, 2003; Casady & Nichols-Larsen, 2004; Debuse, Chandler, &
Gibb, 2005; Hamill, Washington, & White, 2007; Lechner et al., 2007; McGibbon, Benda, Duncan, &
Silkwood-Sherer, 2009; Shurtleff et al., 2009; Silkwood-Sherer, Killian, Long, & Martin, 2012;
Silkwood-Sherer & Warmbier, 2007).
The rationale often cited in support of EAT is the theory that the horse induces in the riders body a
repetitive, cyclic pattern of motion that is similar to that of natural human walking (Beinotti, Correia,
Christofoletti, & Borges, 2010; Benda et al., 2003; Bertoti, 1988; Biery, 1985; Casady & Nichols-Larsen,
2004; Hammer et al., 2005; McGibbon, Andrade, Widener, & Cintas, 1998; Miller, 2007; Quint &
Toomey, 1998; Riede, 1988; Wheeler, 2000). Qualitatively this theory seems reasonable since the rider
sits on the horses back near the pelvis, and the horses pelvis is driven largely by movement of its hind
limbs when walking. Subjectively, simple observation of a person walking and a person riding a horse
reveals that both exhibit translations in, and rotations about, the three principal axes. However,
scientic studies to quantify similarities between walking and riding motions are scant. Numerous
studies have reported on human pelvis kinematics during natural walking (e.g., Lamoreux, 1971;
Thurston & Harris, 1983; Whittle & Levine, 1999; Zhao, Zhang, Wang, & Wang, 2005). Other
studies have reported human pelvis kinematics during horseback riding, but these tend to involve
professional riders during dressage (Mnz, Eckardt, Heipertz-Hengst, Peham, & Witte, 2013), trotting
(Lagarde, Peham, Licka, & Kelso, 2005), or walking on a treadmill (Bystrm, Rhodin, von Peinen,
Weishaupt, & Roepstorff, 2010). To our knowledge, only Fleck (1992) has measured and compared
both walking and riding kinematics in a single study.
Fleck (1992) studied the body motions of 24 healthy children while walking on a treadmill and
while riding on a horse that was walking on an equine treadmill. She reported measurement ranges
in the sagittal and frontal planes of lateral pelvic displacement, lateral pelvic list angle, and the vertical
displacement of the estimated body center of mass. Her results showed similarities between walking
and riding for the pelvic tilt angle, the directions of displacements, and the timing sequences of stride,
but also showed differences in the magnitudes of displacements. The Fleck (1992) study provides an
initial foundation for examining the theory that riding motions are similar to walking motions. However, the study presented only the ranges of a few kinematic measures, and did not look at the temporal characteristics or phase sequencing of those measures. Also, only a single horse was included in
the riding trials, and these trials were performed on an equine treadmill (Fleck, 1992).

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

123

The purpose of the present study was to quantitatively measure and compare human body motions
when walking to those when riding a horse. We sought to expand on previous studies by analyzing
multiple kinematic measures, multiple human subjects, and multiple horses, and by considering temporal characteristics and phase sequencing throughout a gait cycle. The primary focus of this motion
study was the movement patterns of the pelvis, as the pelvis is key for walking (Whittle & Levine,
1999) and is most directly driven by horse movement when riding (Quint & Toomey, 1998). Understanding whether and how riding a horse produces body movements similar to those of human walking may provide insights for improved treatment of humans with neuromuscular disabilities.
2. Methods
A series of video motion capture experiments were performed to measure the three-dimensional
pelvis motions of able-bodied children during walking trials and during riding trials simulating sessions of EAT in a riding arena. A common pool of subjects was used for both the walking trials and
the riding trials. The study was approved at Baylor University by the Institutional Animal Care and
Use Committee (IACUC) and the Institutional Review Board (IRB), and participating children and their
parents gave informed consent.
2.1. Participants
Six able-bodied children three boys and three girls ranging in age from 8 to 12 years were
recruited to participate in this study. The participants had no history of injury or pain in the legs,
hip, back, or neck, and were free of neurological and musculoskeletal impairment. All subjects were
inexperienced riders, with none having ridden in the previous six months. Four horses of varying size
and gait pattern were used, and each were trained for and familiar with the practice of EAT. The horses
included a 14-year old, 13-hand Poa (Bailey, Horse 1), a 12-year old, 15-hand Paint mare (Bonnie,
Horse 2), a 25-year old, 14-hand Quarterhorse mare (Dolly, Horse 3), and an 18-year old, 15-hand Arabian male (Sly, Horse 4).
2.2. Experimental setup
The motion capture experiments were conducted at REACH Therapeutic Riding Center, a Professional Association of Therapeutic Horsemanship (PATH) International certied clinic near McGregor,
Texas, USA, which housed the four horses at the time of the study. Riding trials were recorded in
the centers covered riding arena, which has a surface of arena sand approximately 23  23 m
(75  75 ft). Walking trials were conducted in the centers walkway, which has a smooth, at, concrete
surface approximately 4 m (12 ft) wide. In both locations a camera observation space approximately
2.0  1.4  1.4 m (length  width  height) was calibrated. For both riding and walking trials the
lengthwise dimension of the observation space was aligned with the forward direction of movement,
and was positioned centrally in that direction to provide sufcient room for a straight-line walking
path. The observation space was marked at the corners with brightly-colored plastic cones. A six-camera video motion capture system (SIMI Motion Analysis System, zFlo, Inc., Lexington, MA) was set up
to view the observation space.
The child participants wore snugly tting clothes, an approved riding helmet, and a custom pelvis
belt. The pelvis belt consisted of an elastic hook-and-loop strap with a small aluminum bracket
(100  40  4 mm) secured with screws. The bracket was held by the strap rmly and stably against
the medial posterior pelvis region. A long, slender, aluminum bar (360  20  3 mm) was mounted to
the back of the bracket so that each end extended laterally from the posterior pelvis region. Self-illuminating LED markers were adhesively attached to either end of the bar, and to various points on the
childrens back, neck, helmet, shoulders, and legs (see Fig. 1).
2.3. Experimental protocol
Data was collected during two different sessions on separate days, each involving three subjects.
Before each session the arena sand was raked and smoothed to provide a uniform, level surface. Riding

124

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

Fig. 1. Two child participants showing marker setup for motion capture experiments. The participants wore approved riding
helmets and a custom pelvis belt held by a strap against the medial posterior pelvis region. A slender aluminum bar was
mounted to the belt by a bracket so that each end extended laterally from the posterior pelvis region. Self-illuminating LED
markers were adhesively attached to either end of the bar, and to various points on the childrens back, neck, helmet, shoulders,
and legs.

trials were recorded rst, followed by walking trials. The setup of markers on each subject was
unchanged between riding and walking. For the riding trials each horse was tacked with a blanket
and surcingle that provided a sturdy handle but minimal interference between the movement of
the horse and the riders body. Horses were led by trained clinic staff at a slow walking pace similar
to that typical in the practice of EAT. Each rider rode each horse in turn for at least three recorded
passes through the observation space. One exception is that Sly proved too ill-tempered to continue
participating, so only two subjects rode Sly. Prior to recording, several practice passes were performed
to familiarize horse, rider, and lead walker with the protocol. For walking trials, each participant was
recorded walking through the observation space at least four times at a self-selected, normal walking
pace. Prior to recording, several practice trials were performed to help the participants settle on a
comfortable pace.
2.4. Data processing
The SIMI Motion Capture software was used to track the recorded markers at 60 Hz (30 Hz deinterlaced) for the duration of each pass through the observation space. Coordinates along the anteroposterior (x-axis), superoinferior (y-axis) and mediolateral axis (z-axis) of each tracked marker were
then imported into custom software for visualization and processing. Along the x-axis, the anterior
direction was dened as the positive direction; along the y-axis, the upwards direction was dened
as positive; along the z-axis, the rightward direction was dened as positive. For the scope of this

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

125

study, ve measures were derived from motion of the left and right pelvis bar markers. The average of
these two points was computed to represent the trajectory of the center posterior pelvis. The orientation of the line between these two points was computed to represent pelvic list (in the vertical, frontal
plane, positive leftward) and pelvic twist (in the horizontal, axial plane, positive leftward).
Data processing involved several steps designed to: (1) correct for minor drift of the average forward motion from a straight line aligned with the x-axis; (2) identify a single period of the gait cycle;
(3) smooth out noise; and (4) coerce the data trajectories into periodic form consistent with the cyclic
nature of gait. These steps included: (1) tting a linear regression line to the data in the xz plane and
rotating the data until the regression line aligns with the x-axis; (2) tting a quadratic regression line
to the data in the xz plane and rotating out any quadratic, curved component until the quadratic
regression line becomes straight and aligned with the x-axis; (3) identifying consecutive valleys in
the y-axis data to mark the beginning and end of a single gait cycle; (4) subtracting out the average
forward motion; (5) tting a 9th order Fourier series function to the data trajectories and using the
function ts to smooth and generate periodic, uniformly-populated data sets by which all trials could
be compared in normalized time.

(A)

(C)

(B)

(D)

Fig. 2. Average and standard deviations of stride characteristics for riding (red bars) and walking (green bars). Stride length (A),
velocity (B), period (C), and cadence (D) are shown for each participating horse, all horses, each of three male (M1, M2, M3) and
three female (F1, F2, F3) subjects, the average of all male (All M) and female (All F) subjects, and the average of all subjects (All
Walking). (For interpretation of the references to color in this gure legend, the reader is referred to the web version of this
article.)

126

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

2.5. Data analysis

The processed data was synchronized across trials based on the valley points in the vertical displacement, and normalized in time based on gait cycle period. From the processed data the averages
and standard deviations of stride length, period, cadence, and mean forward velocity of the pelvis center point were computed over each of the following conditions (trial groupings): (1) all subjects riding
each individual horse (Horse 1, Horse 2, Horse 3, Horse 4 in Fig. 2); (2) all trials on all horses (All
Horses); (3) all trials of each subject walking (M1, M2, M3, F1, F2, F3); (4) all male subjects walking
(All M); (5) all female subjects walking (All F); and (6) all subjects walking (All Walking). Averages
and standard deviations were also computed for each of the ve measures (at each corresponding
instant in normalized time) and their ranges over the trial groupings of: (1) each horse; (2) all horses;
and (3) all subjects walking. Finally, the statistical differences across conditions were computed for the
range of each measure using an unpaired Students t-test analysis assuming unequal variance on a
sample, with signicance assumed for p < .05 (Microsoft Excel v12, Microsoft Corp, Redmond, WA).

2.6. Computer animations

The data representing the averages of all trials walking, and of all children riding on each horse,
were used to animate the motion patterns. Custom modeling software was used to generate animations by rendering an adult pelvis (Garner & Pandy, 1999), scaled down to 80% for child size, at the
position and orientation prescribed by each processed data set.
3. Results
3.1. Stride characteristics
The computed results for stride characteristics are shown in Fig. 2. Stride length for the horses
(1.49 0.12 m) was naturally longer on average (p < .001) than that for the children walking
(1.28 0.13 m), although one female subject had an average stride length (F3; 1.38 m) just short of
two of the horses (H1, H4; 1.39 m). Stride velocity was not signicantly different on average
(p = .87) between the horses (1.172 0.11 m/s) and the children (1.166 0.16 m/s), although the childrens velocity was slightly more varied. Stride period was longer on average (p < .001) for the horses
(1.28 0.14 s) than for the children (1.10 0.1 s), and correspondingly, stride cadence was shorter on
average (p < .001) for the horses (47.4 5.4 strides/min) than for the children (54.8 4.7 strides/min).

3.2. Pelvis displacements vs. time

The averages standard deviations of the ve measures (x, y, z, list, and twist) are shown versus
normalized time in Fig. 3 and Fig. 4B. The plots show data averaged over all riding trials and all walking trials.

3.2.1. Vertical (y) displacement

Because each trial was normalized and synchronized according to the valley in vertical displacement, the extremes in Fig. 3A tend to align. The pelvis trajectory exhibits two vertical peaks and
two valleys per cycle both when walking and when riding, corresponding to left and right single limb
phases (of horses hind limbs when riding). The trajectory when walking is fairly sinusoidal, as
expected for bipedal gait. The trajectory for riding is similar, but with small additional undulations
due to the quadrupedal gait. The peaks and valleys for walking have about the same magnitude,
whereas for riding the magnitudes vary slightly, suggesting mild asymmetry. The peaks for walking
are about 25% of the gait cycle separated from the valleys, whereas for riding the drop from peak to
valley covers about 20% of the cycle, and the rise from valley to peak covers about 30% of the cycle.

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

(A) Y

Vertical Displacement (cm)

(C)

Angle (deg)

127

Twist Angle (deg)

Fig. 3. Averages (thick lines) with standard deviations (thin lines) of pelvis displacements versus gait-period-normalized time
for human gait (solid green lines) and when riding (red dashed lines). The time axis is normalized by the period of one gait cycle,
and data over 1.5 gait cycles are shown. The vertical lines at 1.0 show the end of one complete cycle. Variables Y and Z are the
vertical and lateral displacements, respectively. List (L) and Twist (W) angles are angular displacements about the anterior
posterior and vertical axes, respectively. Data from all trials were synchronized by the valleys in the vertical displacement. (For
interpretation of the references to color in this gure legend, the reader is referred to the web version of this article.)

3.2.2. Mediallateral (z) displacement

The lateral displacement pattern, shown in Fig. 3B, reveals a sway to one side and then the other,
occurring once per gait cycle for both walking and riding. The lateral motion for walking remains fairly
sinusoidal and smooth, whereas for riding the evidence of quadrupedal gait is more pronounced with
a small, additional undulation at the peaks on either side. With the data synchronized according to the
vertical displacement, the lateral displacement for riding lags behind that for walking by about a quarter cycle. Thus, for example, the lateral displacement during walking peaks about 25% through the gait
cycle, whereas during riding it peaks closer to 50%.

3.2.3. Anteriorposterior (x) displacement

As with vertical displacement, the anteriorposterior displacement in Fig. 4B consists of dual peaks
and valleys relative to the average forward motion, corresponding to the propulsion phase of each leg
(i.e., hind legs for the horse). The shape of the motion pattern is fairly sinusoidal for both walking and
riding, although some asymmetry is exhibited for riding (the peak for one leg is not equal in magnitude to the other). The peak anterior displacement for riding roughly coincides with peak vertical displacement, but the peak anterior displacement for walking leads peak vertical displacement by about
15% of the gait cycle.

128

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

(A)

(B)

(C)

(D)

Fig. 4. Average pelvis displacements shown spatially (A, C, and D) and versus gait-period-normalized time (B) for human gait
(solid green lines) and when riding (dashed red lines). The thin, solid, colored lines are riding averages for the individual horses.
Variables X, Y, and Z are the anteriorposterior, vertical, and lateral displacements, respectively. The spatial graphs show pelvis
displacements as seen from a top view (A), a back view (C), and a right side view (D). (For interpretation of the references to
color in this gure legend, the reader is referred to the web version of this article.)

3.2.4. Pelvic list (L) angle

As with lateral displacement, the list angle pattern in Fig. 3C reveals a listing to one side and then
the other once per gait cycle for both walking and riding. However, unlike lateral displacement, the list
angle for walking remains essentially in phase with that for riding, both having peaks about 25%
(rightward) and 75% (leftward) through the gait cycle. Interestingly, with list angle the walking data
reveals the additional undulation at each peak whereas riding does not.

3.2.5. Pelvic twist (W) angle

The twist angle pattern, shown in Fig. 3D, also reveals a single peak in each direction for both walking and riding, with both patterns exhibiting a fairly triangular shape. However, twist angle for walking is almost completely out of phase with that for riding. For example, during riding the leftward
twist peaks at around 35% of the gait cycle, whereas for walking it peaks at about 95%.

3.3. Pelvis spatial trajectories

Fig. 4 shows spatial trajectories of the center, posterior pelvis point from the top (A), back (C), and
side (D) orthographic views, arranged so the abscissa scale aligns in A and C, and the ordinate scale
aligns in C and D. Fig. 5 shows various other sample spatial views exhibiting interesting movement

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

129

features, including those with pelvis orientation angles. The top, back, and side spatial trajectories are
also animated in the movie les.
3.3.1. X vs. Z top view
Pelvis displacement from the top view (Fig. 4A) exhibits a lemniscate motion pattern (innity
shape). This shape is very clear for walking, as the pelvis sweeps laterally with anterior displacement,
and then remains lateral during posterior displacement, before reciprocating in the other direction. A
similar pattern occurs during riding, except that the lateral displacement is slightly delayed and more
prolonged, resulting a distortion of the lemniscate crossover point anteriorly, and the wings
posteriorly.
3.3.2. Y vs. Z back view
Pelvis displacement from the back view (Fig. 4C) exhibits a distorted lemniscate motion pattern for
both walking and riding. During walking the pelvis rises superiorly along the lateral wing of the lemniscate, then drops quickly as the pelvis begins to sweep laterally toward the opposite side. The result
is a lemniscate crossover point near the valley of vertical displacement. During riding the pelvis also
rises superiorly along the lateral wing of the lemniscate. However, it tends to remain high through the
lateral sweep until dropping again only as it reaches the opposite side. The result is a lemniscate crossover point near the zenith of vertical displacement.

(A)

(B)

(C)

(D)

Fig. 5. Average pelvis displacements shown spatially for human gait (solid green lines) and when riding (dashed red lines). The
thin, solid, colored lines are riding averages for the individual horses. Variables X, Y, and Z are the anteriorposterior, vertical,
and lateral displacements, respectively. List (L) and Twist (W) angles are angular displacements about the anteriorposterior
and vertical axes, respectively. (For interpretation of the references to color in this gure legend, the reader is referred to the
web version of this article.)

130

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

3.3.3. Y vs. X side view

A side view of pelvis displacement (Fig. 4D) reveals a double-loop motion pattern as the pelvis traverses its anteriorposterior (x) range and its superiorinferior (y) range twice during each gait cycle.
During walking the pelvis drops inferiorly along the posterior edge of the loop, and rises along the
anterior edge, moving anteriorly primarily along the inferior edge, resulting in a counter-clockwise
looping motion as viewed from the right side. During riding, the loop is somewhat compressed along
a near 45-degree line, moving anteriorly and superiorly in unison, then moving posteriorly and

mean & standard deviation

p values (* < 0.05)

Horse 2 Horse 3 Horse 4 Child Gait

X-Range Horse 1
(cm)
Horse 2
Horse 3
Horse 4
All Horses
Child Gait

0.0071 * 0.1075 0.6742

0.0333 * 0.0060 *
0.1210
-

0.0001 *
0.3635
0.0057 *
0.0037 *
0.0018 *
-

Y-Range Horse 1
(cm)
Horse 2
Horse 3
Horse 4
All Horses
Child Gait

0.0001 * 0.8372 0.0019 *

0.0001 * 0.8232
0.0013 *
-

0.0024 *
0.0024 *
0.0095 *
0.0201 *
0.7742
-

Z-Range Horse 1
(cm)
Horse 2
Horse 3
Horse 4
All Horses
Child Gait

0.1127
-

0.0269 *
0.8632
0.9743
0.5195
0.4157
-

L-Range Horse 1
(deg) Horse 2
Horse 3
Horse 4
All Horses
Child Gait

0.0201 * 0.0068 * 0.4212

0.4385 0.0834
0.1515
-

W-Range Horse 1
(deg) Horse 2
Horse 3
Horse 4
All Horses
Child Gait

0.8630
-

0.0062 * 0.0063 *
0.8257 0.4783
0.4492
-

0.1181
0.0623
0.0502
0.9116
0.1956
-

0.0027 * 0.0763 0.0074 *

0.0318 * 0.0750 0.0177 *
0.0056 * 0.2889
0.0012 *
0.0129 *
-

Fig. 6. Averages (thick bars), standard deviations (thin bars), and statistical differences of pelvis displacement ranges for riding
and walking. Variables X, Y, and Z are the anteriorposterior, vertical, and lateral displacements, respectively. List (L) and Twist
(W) angles are angular displacements about the anteriorposterior and vertical axes, respectively. Diamonds mark the range of
averages (see Section 4.2).

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

131

inferiorly in unison. In addition, the pelvis anterior motion occurs on the superior edge of the loop,
resulting in a more attened, clock-wise looping motion as viewed from the right side.

3.3.4. X vs. twist

Fig. 5A shows that during both walking and riding the pelvis moves toward peak twist angle in each
direction as it also moves toward peak anterior displacement. Then, the pelvis twists back toward a
neutral position during posterior displacement.

3.3.5. Z vs. twist

In contrast, Fig. 5B shows that during both walking and riding the pelvis reaches extreme twist
angles in each direction generally as it is passing through the median plane (horizontal axis of
Fig. 5B) of lateral displacement. For both walking and riding the pelvis tends to be twisted leftward
while moving rightward, and rightward while moving leftward (clockwise trajectory in the plotted
data).

3.3.6. Y vs. list

For both walking and riding the pelvis list angle tends to be neutral at the valley of vertical displacement, and to increase in magnitude as the pelvis rises vertically (Fig. 5C). This increase continues
for riding so that peak list angle coincides with peak vertical displacement. For walking the increase
reverses so that list angle approaches neutral again at peak vertical displacement.

3.3.7. List vs. twist

During riding the pelvis lists to one side or the other in near unison as it twists to that side (Fig. 5D).
During walking, however, the list and twist angles are out of phase by about 25% of the gait cycle, so
that twisting toward one side or the other leads listing to that side.

3.4. Pelvis displacement ranges

As shown in Fig. 6, no statistical difference on average was seen between walking and riding for the
ranges of vertical pelvis displacement (p = .77), lateral pelvis displacement (p = .42), and pelvis list
angle (p = .19). However, the ranges for anteriorposterior displacement and twist angle were statistically different (p = .002 and p = .013, respectively) between walking and riding, with values for walking being larger on average than those for riding. In each of the coordinate directions the average range
of pelvis translation for riding was about 4 cm, though the range for riding on individual horses varied
from about 3 cm to over 5 cm. The vertical displacement when riding on each horse was statistically
different on average than when walking, but on Horses 1 and 3 the range was smaller, and on Horses 2
and 4 the range was larger. Of the ve measures, mediallateral displacement exhibited the least statistical variability, with statistical differences only observed in comparison with Horse 1.

4. Discussion
This study was inspired by the practice of EAT, which uses horse movement to treat individuals
with various forms of neuromuscular disability. The purpose was to examine the theory that riding
on a horse produces in the riders body motion patterns that are similar to those of natural human
activities such as walking. Mainstream therapeutic strategies seem to recognize the value of repetitive
motion and the practice of natural activities, but have limited tools for providing such motion. If an
activity such as riding a horse can help generate natural movement patterns that help stimulate the
rider to correct posture and maintain balance, it may prove effective in providing therapeutic benets
that could translate into improved quality of life.

132

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

4.1. Signicance of the study

This study builds on previous work such as that of Fleck (1992), but is novel in several key respects.
First, this study utilizes 3D motion analysis in the comparison of walking and riding motions. Fleck
(1992) collected video data from a back view and a side view independently, and reported this data
in the form of one-dimensional displacement ranges in the directions of vertical, lateral, and listing
motion. We have presented data on ve measures in the form of displacement ranges, sequenced displacement versus time plots, and two-dimensional spatial plots. Second, this study involved the same
pool of multiple human subjects, and multiple animal subjects in the comparison of walking and riding. Fleck (1992) included multiple subjects, but only a single horse. Our study included four different
horses, and showed that there can be substantial variation between the horses. Finally, in this study
we collected data during walking and riding trials performed over ground. Fleck (1992) used a treadmill for both walking and riding trials, which may inuence results. For example, the treadmill was set
to a common, xed speed, which may explain why Fleck (1992) observed similar stride durations for
walking and riding.
4.2. Issues related to analysis
To facilitate comparison of motion patterns across participants and conditions, data processing in
this study included steps of synchronization, normalization, and averaging. The synchronization step
aligned the cyclic data in time based on the timing of valleys in the vertical pelvis displacement. This
approach placed the emphasis on comparison of motion features experienced by the human rider
rather than on relating motion features to specic gait cycle events such as heal strike and toe off,
which do not necessarily correlate between human and equine gait. The normalization step scaled
the time dimension of each trial to the gait cycle period so that the relative timing of motion patterns
could be compared and averaged. As in Lamoreux (1971), data were not normalized by subject
anthropometry.
The averaging step blended multiple trials into a composite representation of the motion pattern.
Averaging of data across trials and across subjects is not uncommon (e.g., Bystrom, Rhodin, von
Peinen, Weishaupt, & Roepstorff, 2010; Lamoreux, 1971; von Peinen et al., 2009), but it can result
in some artifacts and distortions. For example, despite synchronization and normalization, the timing
of peaks and valleys in the data can be slightly out of phase, even for multiple trials involving the same
participant and conditions. This temporal nonalignment of peaks and valleys may attenuate the amplitude in the computed average motion pattern, resulting in, for example, a range of displacement that
is less than that of any of the trials included in the average. For this reason, the range data presented in
Fig. 6 represents the averages of the ranges (range computed rst, and then averaged) as bars, and the
ranges of the averages (averages computed rst) as diamonds (and also revealed in the graphs of
Figs. 35).
4.3. Walking motions compared to previous studies
Comparing the walking data of our study to similar data from other studies gives us condence that
the gait motion patterns exhibited by our subjects are fair representations of natural human gait. With
respect to stride characteristics, the data from our child subjects compare well to data from previous
studies of child gait (e.g., Fleck, 1992; Oberg, Karsznia, & Oberg, 1993). For example, our observed selfselected gait speeds for boys (101.8 17.0 cm/s) fall between the slow (88.7 12.0 cm/s) and normal
(132.3 19.6 cm/s) speeds for boys reported by Oberg et al. (1993). For girls, our self-selected gait
speeds (122.5 11.5 cm/s) fall between the normal (108.6 11.2 cm/s) and fast (146.7 17.6 cm/s)
speeds reported by Oberg et al. (1993). Our step frequencies of 100.2 7.2 steps/min (boys) and
113.4 7.2 steps/min (girls) are similar to those from Oberg et al. (1993), who reported speeds of
100.8 11.4 steps/min for boys walking at a slow pace and 118.2 10.2 steps/min for girls walking
at a normal pace), and Fleck (1992), who reported a speed of 118.8 10.6 steps/min for all subjects.
The pelvic displacement patterns of our participants also compare well with similar data from previous adult gait studies. Lamoreux (1971) studied one adult male walking on a treadmill. Crosbie and

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

133

Vachalathiti (1997) studied 108 adults performing gait trials at a self-selected walking speed over
level oor. Cappozzo (1981) studied 5 similarly-built, young adult males, and Whittle and Levine
(1999) studied 20 adult males walking over level ground at differing speeds. Our average vertical displacement range was 39 mm, which is comparable to Crosbie and Vachalathiti (1997) who reported a
range of 36 mm, and Cappozzo (1981) who reported a range of 45 mm. Our data also shows the same
smooth, sinusoidal, double-peak pattern compared to Crosbie and Vachalathiti (1997) and Cappozzo
(1981). Our lateral displacement range of 45 mm is similar to that of Crosbie and Vachalathiti (1997),
who reported a range of 43 mm, and our data shows the same triangular pattern with fairly constant
speed between peaks. The anteriorposterior displacement range of 43 mm reported by Cappozzo
(1981) is almost identical to ours (44 mm). Crosbie and Vachalathiti (1997) reports the same list angle
range as we observed (6.9), while Whittle and Levine (1999) reports 7.7. Our list angle pattern shows
the same swing to either side, with a slight undulation at the peak. Finally, Crosbie and Vachalathiti
(1997) shows the same triangular twist angle pattern as in our study, but with a range (4.3) much
lower than ours (12.0) and that of Whittle and Levine (1999), who reported a range of 10.4.
4.4. Riding motions compared to previous studies
Although there appears to be no data in the literature suitable for direct comparison, several riding
studies provide data promoting condence that our subjects exhibited fairly typical riding motions.
Two studies measured the motions of the backs of healthy horses at points corresponding to where
a rider would sit: Licka, Peham, and Zohmann (2001) utilized horses walking on a treadmill, and
the protocol for Janura, Dvorakova, Peham, Svoboda, and Elfmark (2010) included horses walking over
ground. These studies report average vertical displacements of 3.2 cm and 3.9 cm, respectively, which
compare well to the 4.0 cm average vertical displacement of rider pelvis that we observed. For lateral
displacements these two horse studies report values of 5.5 cm and 6.1 cm, respectively, which are
comparable to the 4.0 cm average lateral displacement of rider pelvis that we observed, and the
5.6 cm average reported by Fleck (1992).
Several other studies report pelvis orientation angles of experienced riders on trained dressage
horses performing collected walk on a treadmill. Bystrom et al. (2010) and von Peinen et al. (2009)
report the range of average pelvis twist angle to be 6.2 and 7.2, respectively, which compare well
to our value of 5.7. Bystrom et al. (2010) also reports the average of pelvis twist angle ranges as
8.2, which is identical to our value. These studies also report the range of average pelvis list angle
to be 4.0 and 4.3, respectively, which is a bit less than our value of 6.5, and a value of 10 reported
by Fleck (1992). A 5.6 average of pelvis list angle ranges reported by Bystrom et al. (2010) is also a bit
less than our average value of 7.9.
4.5. Variations within walkers, horses, and riders
In comparing walking to riding it is helpful to keep in mind that the results show fairly substantial
variation across subjects within walking, across different horses, and across riders on individual
horses. These variations are illustrated in Fig. 6 by, respectively, standard deviations for child gait, differences in ranges for the various horses, and standard deviations for each horse. For example, the
standard deviation for twist angle over all subjects was about 50% of the full range. The vertical displacement range varied from about 3.0 cm for two horses, to about 5.5 cm for two other horses. And,
the standard deviation for anterio-posterior (x) displacement over all riders on Horse 2 was about 30%
of the full range. Thus, the data indicates a fairly broad pattern of riding motion for comparison with a
fairly broad pattern of walking motion.
4.6. Similarities between walking and riding
Despite variations across walkers, horses, and riders, the results of this study indicate that there are
strong similarities between the human pelvis motions when walking and when riding on a horse.
For example, the average pelvis displacement ranges presented in Fig. 6 reveal that the general
amplitudes of motion for riding and walking are quite similar. For four of the ve measures, including

134

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

anteroposterior (X) displacement, superoinferior (Y) displacement, mediolateral (Z) displacement, and
list angle (L), there were either no statistical differences between riding and walking, or the walking
averages were within the spread of riding averages on the different horses (i.e., green bars in Fig. 6 fall
between the minimum and maximum red bars). Although there was a statistical difference for twist
angle (W) on average across all horses, there was no statistical difference observed between walking
and riding on Horse 3. In short, there seems to be at least as much variability in general motion
amplitude between riding on the different horses as there is variability between the walking and
riding data. For example, nearly half (14 of 30) of the p-values in the Horse 2 through Horse 4 columns
of Fig. 6 indicate signicant differences between horses, whereas 12 of 30 p-values in the child gait
column indicate signicant differences between riding and walking. Also, for every measure there
can be found two horses whose average ranges differ by more than the difference between human
walking and all horses, and also more than between human walking and at least one individual horse.
For example, for twist angle (W) the Horse 3 average is about 5 greater than the Horse 4 average,
whereas the walking average is only about 4 greater than the average of all horses, and only about
2 greater than the Horse 3 average.
In addition to amplitudes, the general patterns of motion for riding also exhibit many similarities
with those of walking. For both riding and walking the forward and vertical displacement patterns
exhibit fairly sinusoidal shapes with double peaks and double valleys each cycle (Figs. 3A and 4B).
Likewise, the lateral, list, and twist displacements exhibit similar patterns of single major peaks and
single major valleys each cycle (Fig. 3BD). In all of these cases the general duration of time between
peaks and valleys is similar for riding and walking. The spatial views of Figs. 4 and 5 provide additional
examples. The top view (Fig. 4A) shows that for both activities the pelvis tends to move laterally and
anteriorly across the center position, and then move posteriorly out at the lateral positions. The back
view (Fig. 4C) shows that for both activities (though a little more so for riding) the pelvis tends to
move vertically when displaced laterally, and then to traverse to the opposite side with relatively little
vertical movement. The side view (Fig. 4D) reveals a pelvis trajectory that makes two loops each gait
cycle, though the loop for walking is more circular. Plots A and B of Fig. 5 show that for both riding and
walking the pelvis tends to twist internally (facing inward toward the center line) as it traverses from
a posterior, lateral position on one side across to an anterior, lateral position on the other side, and
then straighten out (facing forward) as it sweeps back posteriorly. Fig. 5C also shows that for both riding and walking the list angle tends to reach a neutral orientation at the lowest vertical position, and
then lean laterally as it moves upward.
4.7. Differences between walking and riding
Along with the similarities there were also some distinct differences observed between the motion
patterns of the human pelvis during walking and during riding. The vertical displacement range for
walking was statistically different from that of each of the individual horses, despite the fact that,
as noted above, there was no statistical difference between walking and the average over all horses.
And, as noted above, the range of pelvis twist angle for walking was statistically different from, and
on average higher than, that for the horses. There was, however, a fairly large standard deviation
for the walking twist range, with some children averaging within the variability of riding, and others
well above it.
The results also reveal some qualitative differences in motion patterns between walking and riding.
For example, the lateral translation (Fig. 3B) for riding (red dashed line) reveals an additional undulation at each major peak and valley that is not seen in walking (green solid line). This undulation is
likely due to the mechanics of the quadrupedal gait. Conversely, the list angle (Fig. 3C) for walking
(green solid line) reveals an additional undulation at each major peak that is not seen in riding (red
dashed line). This undulation occurs in walking as the pelvis lists toward the stepping leg, then
momentarily levels out again as the opposite leg swings through, before listing again into the step.
Perhaps the most substantial differences observed between walking and riding relate to the timing
of some motion features. When, as in this study, the timing of peaks and valleys in vertical displacement were aligned (as in Fig. 3A), then the peaks and valleys of lateral displacement (Fig. 3B), forward
displacement (Fig. 4B), and twist angle (Fig. 3D) do not align. These phase differences become

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

135

apparent in the spatial views. For example, in riding the forward and upward displacements occur
concurrently, resulting in a side view trajectory that is more attened, angled along a diagonal,
45-degree line (Fig. 4D, thick dashed red line). In walking, the pelvis moves upwards as it is peaking
forwards, and then moves backwards as it peaks over the top, resulting in a more circular side view
trajectory (Fig. 4D, thick solid green line). The back view (Fig. 4C) provides another example. For riding
(thick dashed red line), the pelvis tends to move laterally while peaking at its high point vertically, and
then remains lateral while it drops and rises back up. For walking (thick solid green line), the pelvis
tends to move laterally during both the drop and rise, reaching the extreme lateral positions about
the same time as it peaks vertically.
4.8. Applications and future work
We expect that this study will help increase understanding about pelvis motion patterns when
walking and when riding on a horse, and about the similarities and differences between walking
and riding motions. We further hope that this study will help increase understanding about how
motion patterns may play a role in therapeutic strategies such as EAT, and how such therapies may
be improved by proper selection of motion patterns. It remains for future work to (1) quantify specic
benets of such motion therapies, (2) determine whether benets require complex motion patterns or
whether basic displacement patterns may sufce, and (3) investigate the role of motion for therapeutic benets compared to other factors including interaction with live animals, warmth from an
animals body heat, stretching due to sitting on an animal, moving forward through space, and being
outdoors. Although the focus of this work is on kinematics, additional insights may also be gained in
future work comparing the dynamics of walking and riding.
4.9. Limitations of the study
We anticipate future studies to improve upon the current study in several ways. For example, gait
speed in the current study was not carefully controlled for either the human walking trials or the
horse riding trials. The human subjects walking were asked to choose a preferred speed, and the horse
leads were asked to simulate a normal therapy session. These conditions reect what may naturally
occur. However, some past studies (e.g., Lamoreux, 1971) have shown that gait speed affects motion
patterns. Future studies that more tightly control gait speed, and consider different speeds, may provide additional insights and yield less variability in the results. Also, the current study focused solely
on motion of the pelvis. Because an anterior pelvis marker tended to get obscured during the riding
trials of this study, pelvic tilt angle in the sagittal plane was not reported. Future studies of walking
and riding may include other kinematic measures such as pelvic tilt angle, and spine, neck, and shoulder angles. Finally, this study included typically three trials each of six able-bodied children and four
horses. Future studies could expand on this participant set with additional children, additional horses,
additional age ranges including adults, and individuals with disabilities.
5. Conclusion
This study presents novel motion capture data relating human pelvis motions while walking and
while riding a horse. If an activity such as riding can reproduce natural human movement patterns
for the rider, it may provide therapeutic benets to individuals with disabilities who cannot move naturally for themselves. The motion patterns observed in this study were consistent with previous studies of walking, and with previous studies of riding. There was substantial variation in riding motion
patterns across different horses, and in most respects these enveloped the motion patterns of human
walking. Displacement amplitudes during walking were similar to those of riding for pelvic list angle
and anteroposterior, superoinferior, and mediolateral translation. Key features of the pelvic motion
trajectories, such as the number and general shape of peaks and valleys, were similar for walking
and riding. In the case of some key features the pelvic motion trajectories for riding exhibited
additional undulations, and the timing was somewhat out of phase compared to walking.

136

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

Follow-up studies should focus on expanded participant populations and kinematic measures,
including pelvic tilt angle in the sagittal plane.
Acknowledgments
We gratefully acknowledge the support and internal funding provided by Baylor University, staff,
and students. We also appreciate the assistance provided by the following colleagues: Larry Barnett,
director of the REACH Therapeutic Riding Center in McGregor, Texas; Shirley Wills, director of the
Heart of Texas Therapeutic Riding Center in West, Texas; Nancy Krenek, director of the Ride On Center
for Kids (R.O.C.K.) in Georgetown, Texas; Beth Lanning, associate professor of Health, Human Performance and Recreation, Baylor University in Waco, Texas; and Ken Alford, department head of Health
and Human Performance, Texas A&M-Commerce, in Commerce, Texas.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, in the online version, at http://
dx.doi.org/10.1016/j.humov.2014.06.011.
References
American Hippotherapy Association (2010). Hippotherapy as a treatment strategy. Retrieved from <http://
www.americanhippotherapyassociation.org/hippotherapy/hippotherapy-as-a-treatment-strategy/>.
Beinotti, F., Correia, N., Christofoletti, G., & Borges, G. (2010). Use of hippotherapy in gait training for hemiparetic stroke.
Arquivos de Neuro-Psiquiatria, 68, 908913.
Benda, W., McGibbon, N. H., & Grant, K. L. (2003). Improvements in muscle symmetry in children with cerebral palsy after
equine-assisted therapy (hippotherapy). The Journal of Alternative and Complementary Medicine, 9, 817825.
Bertoti, D. (1988). Effect of therapeutic horseback riding in children with cerebral palsy. Physical Therapy, 68, 15051512.
Biery, M. J. (1985). Riding and the handicapped. Veterinary Clinics of North America Small Animal Practice, 15, 345354.
Bystrm, A., Rhodin, M., von Peinen, K., Weishaupt, M. A., & Roepstorff, L. (2010). Kinematics of saddle and rider in high-level
dressage horses performing collected walk on a treadmill. Equine Veterinary Journal, 42, 340345.
Casady, R. L., & Nichols-Larsen, D. S. (2004). The effect of hippotherapy on ten children with cerebral palsy. Pediatric Physical
Therapy, 16, 165172.
Cappozzo, A. (1981). Analysis of the linear displacement of the head and trunk during walking at different speeds. Journal of
Biomechanics, 14, 411425.
Crosbie, J., & Vachalathiti, R. (1997). Synchrony of pelvic and hip joint motion during walking. Gait & Posture, 6, 237248.
Damiano, D. L. (2006). Activity, activity, activity: Rethinking our physical therapy approach to cerebral palsy. Physical Therapy,
86, 15341540.
Debuse, D., Chandler, C., & Gibb, C. (2005). An exploration of German and British physiotherapists views on the effects of
hippotherapy and their measurement. Physiotherapy Theory and Practice, 21, 219242.
Dodd, K. J., Taylor, N. F., & Damiano, D. L. (2002). Systemic review of strengthening for individuals with cerebral palsy. Archives of
Physical Medicine and Rehabilitation, 83, 11571164.
Fleck, C. A. (1992). Hippotherapy: Mechanics of human walking and horseback riding. In B. T. Engel (Ed.), Rehabilitation with the
aid of the horse: A collection of studies (pp. 153176). Durango, CO: Barbara Engel Therapy Services.
Garner, B. A., & Pandy, M. G. (1999). A kinematic model of the upper limb based on the Visible Human Project (VHP) image
dataset. Computer Methods in Biomechanics and Biomedical Engineering, 2, 107124.
Geiger, R. A., Allen, J. B., OKeefe, J., & Hicks, R. R. (2001). Balance and mobility following stroke: Effects of physical therapy
interventions with and without biofeedback/forceplate training. Physical Therapy, 81, 9951005.
Hamill, D., Washington, K. A., & White, O. R. (2007). The effect of hippotherapy on postural control in sitting for children with
cerebral palsy. Physical & Occupational Therapy in Pediatrics, 27, 2342.
Hammer, A., Nilsagard, Y., Forsberg, A., Pepa, H., Skargren, E., & Oberg, B. (2005). Evaluation of therapeutic riding (Sweden)/
hippotherapy (United States). A single-subject experimental design study replicated in eleven patients with multiple
sclerosis. Physiotherapy Theory and Practice, 21, 5177.
Hogan, N., Krebs, H. I., Rohrer, B., Palazzolo, J. J., Dipietro, L., Fasoli, S. E., et al (2006). Motions or muscles? Some behavioral
factors underlying robotic assistance of motor recovery. Journal of Rehabilitation Research and Development, 43, 605618.
Janura, M., Dvorakova, T., Peham, C., Svoboda, Z., & Elfmark, M. (2010). The inuence of walking speed on equine back motion in
relation to hippotherapy. Wiener Tierarztliche Monatsschrift, 97, 8791.
Ketelaar, M., Vermeer, A., Hart, H., van Petegem-van Beek, E., & Helders, P. J. (2001). Effects of a functional therapy program on
motor abilities of children with cerebral palsy. Physical Therapy, 81, 15341545.
Lagarde, J., Peham, C., Licka, T., & Kelso, J. A. S. (2005). Coordination dynamics of the horserider system. Journal of Motor
Behavior, 37, 418424.
Lamoreux, L. W. (1971). Kinematic measurements in the study of human walking. Bulletin of Prosthetics Research, 10, 384.
Lechner, H. E., Kakebeeke, T. H., Hegemann, D., & Baumberger, M. (2007). The effect of hippotherapy on spasticity and on mental
well-being of persons with spinal cord injury. Archives of Physical Medicine and Rehabilitation, 88, 12411248.

B.A. Garner, B.R. Rigby / Human Movement Science 39 (2015) 121137

137

Licka, T. F., Peham, C., & Zohmann, E. (2001). Treadmill study of the range of back movement at the walk in horses without back
pain. American Journal of Veterinary Research, 62, 11731179.
McGibbon, N. H., Andrade, C.-K., Widener, G., & Cintas, H. L. (1998). Effect of an equine-movement therapy on gait, energy
expenditure, and motor function in children with spastic cerebral palsy: A pilot study. Developmental Medicine and Child
Neurology, 40, 754762.
McGibbon, N. H., Benda, W., Duncan, B. R., & Silkwood-Sherer, D. (2009). Immediate and long-term effects of hippotherapy on
symmetry of adductor muscle activity and functional ability in children with spastic cerebral palsy. Archives of Physical
Medicine and Rehabilitation, 90, 966974.
Miller, F. (2007). Physical therapy of cerebral palsy. New York, NY: Springer.
Miller, E. L., Murray, L., Richards, L., Zorowitz, R. D., Bakas, T., Clark, P., et al (2010). Comprehensive overview of nursing and
interdisciplinary rehabilitation care of the stroke patient: A scientic statement from the American Heart Association.
Stroke, 41, 24022448.
Mnz, A., Eckardt, F., Heipertz-Hengst, C., Peham, C., & Witte, K. (2013). A preliminary study of the inertial sensor-based method
for the assessment of human pelvis kinematics in dressage riding. Journal of Equine Veterinary Science, 33, 950955.
Oberg, T., Karsznia, A., & Oberg, K. (1993). Basic gait parameters: Reference data for normal subjects, 1079 years of age. Journal
of Rehabilitation Research and Development, 30, 210223.
Quint, C., & Toomey, M. (1998). Powered saddle and pelvic mobility: An investigation into the effects on pelvic mobility of
children with cerebral palsy of a powered saddle which imitates the movements of a walking horse. Physiotherapy, 84,
376384.
Riede, D. (1988). Physiotherapy on the horse. Riderwood, MD: Therapeutic Riding Services.
Shurtleff, T. L., Standeven, J. W., & Engsberg, J. R. (2009). Changes in dynamic trunk/head stability and functional reach after
hippotherapy. Archives of Physical Medicine and Rehabilitation, 90, 11851195.
Silkwood-Sherer, D., & Warmbier, H. (2007). Effects of hippotherapy on postural stability in persons with multiple sclerosis.
Journal of Neurologic Physical Therapy, 31, 7784.
Silkwood-Sherer, D. J., Killian, C. B., Long, T. M., & Martin, K. S. (2012). Hippotherapy An intervention to habilitate balance
decits in children with movement disorders: A clinical trial. Physical Therapy, 92, 707717.
Stanger, M., & Oresic, S. (2003). Rehabilitation approaches for children with cerebral palsy: Overview. Journal of Child Neurology,
18, S79S88.
Tecklin, J. S. (2007). Pediatric physical therapy. Philadelphia, PA: Lippincott Williams & Wilkins.
Thurston, A. J., & Harris, J. D. (1983). Normal kinematics of the lumbar spine and pelvis. Spine, 8, 199205.
Veneman, J. F., Kruidhof, R., Hekman, E. E. G., Ekkelenkamp, R., van Asseldonk, E. H. F., & van der Kooij, H. (2007). Design and
evaluation of the LOPES exoskeleton robot for interactive gait rehabilitation. IEEE Transactions on Neural Systems and
Rehabilitation Engineering, 15, 379386.
von Peinen, K., Wiestner, T., Bogisch, S., Roepstorff, L., Van Weeren, P. R., & Weishaupt, M. A. (2009). Relationship between the
forces acting on the horses back and the movements of rider and horse while walking on a treadmill. Equine Veterinary
Journal, 41, 285291.
Wheeler, A. (2000). Hippotherapy as a specic treatment: A review of literature. In B. T. Engel (Ed.), Therapeutic riding II:
Strategies for rehabilitation (pp. 2529). Durango, CO: Barbara Engel Therapy Services.
Whittle, M. W., & Levine, D. (1999). Three-dimensional relationships between the movements of the pelvis and lumbar spine
during normal gait. Human Movement Science, 18, 681692.
Willoughby, K. L., Dodd, K. J., Shields, N., & Foley, S. (2010). Efcacy of partial body weight-supported treadmill training
compared with overground walking practice for children with cerebral palsy: A randomized controlled trial. Archives of
Physical Medicine and Rehabilitation, 91, 333339.
Zhao, L., Zhang, L., Wang, L., & Wang, J. (2005). Three-dimensional motion of the pelvis during human walking. In IEEE
international conference (Vol. 1, pp. 335339).