Sports Biomechanics, 2014

Vol. 13, No. 3, 215–229, http://dx.doi.org/10.1080/14763141.2014.933580

Generation of vertical angular momentum in single,

double, and triple-turn pirouette en dehors in ballet

JEMIN KIM1, MARGARET A. WILSON2, KUNAL SINGHAL3,

SARAH GAMBLIN4, CHA-YOUNG SUH5, and YOUNG-HOO KWON1
1
Biomechanics Laboratory, Texas Woman’s University, Denton, TX, USA, 2Department of Theatre and
Dance, University of Wyoming, Laramie, WY, USA, 3Osteopathic and Manipulative Medicine, University
of North Texas Health Science Center, Fort Worth, TX, USA, 4Department of Dance, Texas Woman’s
University, Denton, TX, USA, and 5Department of Dance, Sejong University, Seoul, Republic of Korea

(Received 19 December 2013; accepted 14 May 2014)

Abstract
The purpose of this study was to investigate the vertical angular momentum generation strategies used
by skilled ballet dancers in pirouette en dehors. Select kinematic parameters of the pirouette preparation
(stance depth, vertical center-of-mass motion range, initial shoulder line position, shoulder line angular
displacement, and maximum trunk twist angle) along with vertical angular momentum parameters
during the turn (maximum momentums of the whole body and body parts, and duration and rate of
generation) were obtained from nine skilled collegiate ballet dancers through a three-dimensional
motion analysis and compared among three turn conditions (single, double, and triple). A one-way
(‘turn’) multivariate analysis of variance of the kinematic parameters and angular momentum
parameters of the whole body and a two-way analysis of variance (‘turn’ £ ‘body’) of the maximum
angular momentums of the body parts were conducted. Significant ‘turn’ effects were observed in the
kinematic/angular momentum parameters (both the preparation and the turn) (p , 0.05). As the
number of turns increased, skilled dancers generated larger vertical angular momentums by
predominantly increasing the rate of momentum generation using rotation of the upper trunk and
arms. The trail (closing) arm showed the largest contribution to whole-body angular momentum
followed by the lead arm.

Keywords: Kinematic, rate of generation, motion range, maximum momentum, lead and trail arm,
preparation

Introduction
Turning motions found in all dance forms develop intrinsically from various techniques and
styles of each genre. Turns in classical ballet are called pirouettes in which dancers balance on
one limb (support leg) and perform single or multiple revolutions of the body with the other
limb (gesture leg) rotating around the pivot axis provided by the support foot (Figure 1).
Pirouettes can be performed in two different ways with regard to the direction of rotation of

Correspondence: Young-Hoo Kwon, Biomechanics Laboratory, Texas Woman’s University, P. O. Box 425647, Denton, TX 76204-
5647, USA, Email: ykwon@twu.edu

q 2014 Taylor & Francis

216 J. Kim et al.

Figure 1. Motion sequence of pirouette en dehors used in this study: (A) start position (fifth foot position), (B) retiré
passé, (C) beginning of double-stance; fourth foot position, (D) beginning of angular momentum generation, (E) à la
seconde, (F) gesture foot off, (G) retiré passé, and (H) gesture foot contact. Note that the dancer is turning right so the
right leg is the gesture leg and right arm is the lead arm. A– C is the preparatory motion. C–D is the counter (left)
rotation and some dancers may not show this initial counter rotation. The dancer must generate a sufficient
longitudinal angular momentum during the double-stance phase (C–F) to secure stable turns during the single-
stance phase (F–H).

the gesture leg: en dehors (outward; Figure 1) and en dedans (inward). Traditional
preparations for pirouettes come from a fourth (open position) or fifth (closed position)
(Figure 1) but preparations vary by stylistic and choreographic expectations and dancers
tend to prefer the preparation style they have practiced the longest (Biringen, 2010). From
the preparation of the turn in an open stance, the dancer quickly brings the back (gesture) leg
to the retiré position, with the toes of the gesture leg in front or directly to the side of the
support leg knee and the arms to a closed circle position in front of the waist (en avant).
Angular momentum is generated when the dancer applies a torque about the vertical axis
by pushing sideways in opposite directions with both feet (Laws, 1984). The magnitude of
torque is determined by the magnitudes of the horizontal forces, lengths of the moment
arms, duration of force exertion, and the distance of the center-of-mass (CM) of the body to
the center of rotation (Laws, 2002). The rate of rotation is determined by the moment of
inertia (MOI) of the body about the axis of rotation with the dancer in balance (Laws, 1998).
Once the pirouette is initiated, the loss of angular momentum must be kept minimal to allow
the dancer to execute the desired number of revolutions in the turn. Therefore, it is
important to generate a sufficient vertical angular momentum during the double-stance
 Angular momentum in pirouette en dehors 217

phase (i.e. when both feet are in contact with the ground) and maintain it as much as possible
during the single-stance phase (i.e. when only the support foot is in contact with the ground).
The rate of angular momentum loss during the single-stance phase is a function of the
friction coefficient between the surface and the shoe the dancer is wearing (Laws, 2002).
Sugano and Laws (2002) identified fourth position as ideal for creating angular
momentum as the feet exert equal horizontal forces in opposite directions to produce a
twisting effect. They also reported that in the preparation for successful turns, approximately
60% of the dancer’s weight was over the front foot, necessitating a ‘wider’ fourth, or allongé
position. Biringen (2010), however, found that the best turns were performed from a stance
of 40% of the dancers’ leg length and the position of maximum torque did not correspond to
the position for the best turn, which suggests that the stance has more to do with a dancer’s
preference, rather than optimization of mechanical advantage.
In a study looking at the effect of a ‘wind-up’ on the success of the turn versus an en bloc
strategy where the shoulder and hip stay aligned, Golomer, Toussaint, Bouillette, and Keller
(2009) found that for pirouette en dehors, when standing on the left leg and turning clockwise,
dancers turned shoulders before hips at the beginning of the turn, but employed an en bloc
strategy when standing on the right and turning counterclockwise. They also noted that
when the shoulder/hip angle increased, the success of the turn decreased. Sugano (2010)
suggested that for turns requiring multiple revolutions (. 3) both the arms and the legs
should be lowered, effectively bringing the dancer’s mass closer to the axis of rotation to
decrease the MOI.
Although several biomechanical studies have been conducted on pirouettes en dehors, the
focus of these studies were mainly preparatory stance (Sugano & Laws, 2002), axis of
rotation (Biringen, 2010), and toppling (Lott & Laws, 2012). While these studies point out
that vertical angular momentum is generated from the torque applied by the feet through the
foot –ground interaction during double-stance, what essentially enables the foot – ground
interaction is the dancer’s whole-body motion. In other words, the dancer controls the foot –
ground interaction by manipulating movements in the legs, trunk, shoulders, arms, and
head. Therefore, a comprehensive analysis of the generation of angular momentum must
focus on the dancer’s movement strategies. These strategies may include timing of the
preparation for the turn, action of the body parts completing the position in the turn and the
contribution of open chain movements in the arms, shoulders, and head.
An effective way to gain insights into the angular momentum generation strategies
employed by ballet dancers is to use multiple turn conditions systematically (i.e. single,
double, and triple turns). Dancers must generate more angular momentum as the demand
(number of turns) increases and investigating the movement adjustments made in this
process may provide an in-depth understanding of the overall angular momentum
generation strategies employed by ballet dancers. The purpose of this study, therefore, was to
investigate the vertical angular momentum generation strategies used by skilled ballet
dancers in pirouette en dehors with single, double, and triple turns. It was hypothesized that (1)
skilled dancers would generate larger vertical angular momentum as the number of turns
increases, which would be achieved by longer and higher rate of momentum generation; (2)
dancers would alter kinematic movement patterns of the body as the number of turns
increases; and (3) body parts would exhibit different levels of contribution to the whole-body
angular momentum and the contribution profile would change as the number of turns
increases.
218 J. Kim et al.

Methods
Participants
Eight female and one male skilled collegiate classic ballet dancers were recruited for this
study. The mean (^ SD) mass, height, age, and dance experience were 60.4 ^ 12.1 kg,
161.3 ^ 8.1 cm, 23.0 ^ 2.9 years, and 13.9 ^ 6.2 years, respectively. All participants were
trained in classic ballet and were able to consistently perform triple-turn pirouette en dehors
(turning outward). The dancers were free of major injuries that might hinder the
performance at the time of data collection. The study protocol was approved by the
Institutional Review Board of Texas Woman’s University and all participants provided their
informed consent prior to participation.

Trial conditions
Each participant was required to perform three successful pirouette en dehors trials in each of
the following three turn conditions: single, double, and triple turn. Regardless of their limb
preference, all participants were asked to turn right (clockwise) using right leg as gesture leg
and right arm as lead arm (Figure 1). Pirouette trials were performed on demi pointe (on the
ball of the foot) in ballet technique shoes. The testing movement consisted of a preparation
and the actual turn. In all conditions, the preparatory motion started with the feet in fifth
position with legs externally rotated from the hips and the toes of the gesture foot placed in
front the heel of the support (left) foot (Figure 1A). The dancer executes a demi plié (hip and
knee flexion with heels on the ground) followed by a relevé (rising from the start position to
balance on the support foot) and retiré passé (toe of the gesture foot touching the knee of the
support leg; Figure 1B), and ended with a fourth position with the gesture foot placed behind
the support foot (Figure 1C). The pirouette began from the fourth position with both feet on
the ground (Figure 1D). The dancer opened the lead arm to the side (à la seconde) as they
descended into demi plié (Figure 1E). From this position of hip and knee flexion, the rotation
and weight shift onto the support leg and release of the gesture leg began. At the same time,
the dancer brought the arms together en avant (i.e. arms in front of the trunk at the height of
xyphoid process) as the gesture leg returned to retiré devant (i.e. gesture toe touching the
front of the support knee) and the body began the clockwise revolution on the stance leg
(Figure 1F and G). The turn ended as the dancer’s facing returned to the ‘front’ and the
gesture leg returned to the fourth position behind the stance leg (Figure 1H).
All trials were restricted to the en avant arm position and retiré passé gesture leg position
(Figure 1G). A trial with hopping on the support foot, falling out of the turn and/or absence
of clear head spotting (i.e. keeping eyes fixated on a spot) was considered unsuccessful. For
standardization, each trial was performed to the same music (Douglas Shultz, Leaning on
Tradition, Volume 1, Track 34 Pirouette Waltz; 3/4; 128 beats per minute). Dancers were
asked to warm up for at least 10 minutes prior to data collection and were encouraged to
keep moving between trials.

Experimental setup
A 250-Hz 10-camera VICON motion capture system (Model MX T-10; VICON,
Centennial, CO, USA) was used in an indoor motion analysis facility to capture three-
dimensional motion trajectories of the retro-reflective markers (10 mm in diameter) placed
on dancer’s body. The ‘TWUDancer’ marker set with 48 markers was used for motion
capture (Table I). A static T-pose trial (Figure 2A) was conducted to locate joint centers and
 Angular momentum in pirouette en dehors 219

Table I. ‘TWUDancer’ marker set and computed points

Segment Category Count Description

Head Markers 4 Right- and left-head, forehead, and vertex

Computed 1 Head center (mid-point of the right- and left-head markers)
Trunk Markers 5 Right, and left acromions, supra-sternal notch, 7th cervical
vertebra, and 12th thoracic vertebra
Computed 1 Mid-shoulder (mid-point of the shoulder joint centers)
Arms Marker 8£2 Anterior shoulder, posterior shoulder, medial epicondyle,
lateral epicondyle, radial styloid, ulnar styloid, and distal heads
of the 2nd and 5th metacarpals
Computed 4£2 Shoulder (mid-point of the anterior- and posterior-shoulder
markers), elbow (mid-point of the medial and lateral
epicondyle markers), and wrist (mid-point of the radial and
ulnar styloid markers) joints and hand center (mid-point of the
metacarpal markers)
Pelvis Markers 5 Right anterior-superior iliac spine (ASIS), left ASIS, right
ilium, left ilium, and sacrum
Computed 5 Mid-ASIS (mid-point of the ASIS markers), L4/5 (MacKinnon
& Winter, 1993), hip joints (Bell et al., 1990), and mid-
hip (mid-point of the hip joint centers)
Legs Markers 9£2 Greater trochanter,a lateral thigh, medial epicondyle,a lateral
epicondyle, lateral shank, medial malleolus,a lateral malleolus,
toe, and heel
Computed 2£2 Knee (mid-point of the medial and lateral epicondyles) and
ankle (mid-point of the medial and lateral malleoli) joints
a
Used in the static trial only and removed in the motion trials.

a group of markers were removed in the motion trials (Figure 2B) to ensure interference-free
turn of the body (see Table I for details). All participants were asked to wear black leotard,
black swimming cap, and ballet technique shoes.
Two AMTI force plates (Model OR6; Advanced Mechanical Technology, Inc.,
Watertown, MA, USA) were used to collect the ground reaction force data at a sampling
frequency of 250 Hz. Each force plate was covered with Marley dance floor, a heavy-duty
slip-resistant vinyl floor covering used in dance studios and theatre stages. Both feet were on
the front plate at the start position (Figure 1A) but the gesture foot landed on the rear plate at
the end of the preparatory motion and at the end of the turn (Figure 1C and H). The z-axis
of the laboratory reference frame was aligned vertically upward and dancers faced in the
direction of positive y-axis at the beginning of each motion trial (Figure 1). The axes of the
plates were aligned with those of the laboratory reference frame.

Data processing and analysis

Captured motion trials were imported into Kwon3D motion analysis software (Version XP;
Visol, Inc., Seoul, Korea). Three-dimensional trajectory data were filtered by a Butterworth
low-pass filter (fourth-order zero phase lag; 6-Hz cutoff frequency). Dancer’s body was
modeled as a system of 15 linked segments: head, thorax-abdomen, pelvis, upper arms,
forearms, hands, thighs, shanks, and feet (Figure 2 and Table II). The hip joint center was
located using the Tylkowski – Andriacchi hybrid method (Bell, Pedersen, & Brand, 1990)
and the joint between the fourth and fifth lumbar vertebrae (L4/5) was located following
MacKinnon and Winter (1993). All other joints (knees, ankles, shoulders, elbows, and
220 J. Kim et al.

Figure 2. The ‘TWUDancer’ body model: (A) static trial and (B) motion trial. Forty-eight markers were used: 4
head markers, 5 trunk markers, 5 pelvis markers, 16 arm markers, and 18 leg markers. Six leg markers were removed
in the motion trials. Nineteen additional points (including 14 joints) were computed based on the markers. See
Table I for details of the markers/points used. Fifteen segments were defined using the joints and end points.

wrists) were defined respectively as the mid-point of two markers (e.g. anterior and posterior
shoulder markers for the shoulder joint) (Table I). Segmental reference frames were defined
from the markers and computed points for subsequent kinematic and kinetic computations
(see Table II for details). The x-, y-, and z-axis of the segmental frames were aligned with the
mediolateral, anteroposterior, and longitudinal axes of the segments, respectively. The body
segment parameters reported by de Leva (1996) were employed in computing the CM
positions, masses, and principal MOI of the segments.
Orientation matrices of the segments were built from axis unit vectors of the segmental
reference frames (Kwon, 2008):

2 3 2 3
ii x11 x12 x13
6 7 6 7
Ti ¼ 6 j 7 6x
4 i 5 ¼ 4 21
x22 x23 7
5 ð1Þ
ki x31 x32 x33

where i is the segment, T is the 3 £ 3 orientation matrix, i, j, and k are the unit vectors of the
segmental reference frame, and x11 –x33 are the components of the axis unit vectors. Relative
orientation matrices of the segments to their respective linked proximal segments were
 Angular momentum in pirouette en dehors 221

Table II. Definitions of the segmental reference frames

Segment Anatomical plane Axes Linked proximal

segment

Pelvis Transverse X: left ASIS marker to right ASIS marker Global

Y: sacrum marker to mid-ASIS point
Thorax–abdomen Frontal Z: mid-ilium point to mid-shoulder point Pelvis
Y: C7 marker to SN marker
Head Frontal Z: head center to vertex marker Thorax– abdomen
X: left-head marker to right-head marker
Right upper arm Frontal Z: elbow joint to shoulder joint Thorax– abdomen
X: medial epicondyle to lateral epicondyle
Right forearm Frontal Z: wrist joint to elbow joint Right upper arm
X: ulnar styloid marker to radial styloid marker
Right hand Frontal Z: hand center point to wrist joint Right forearm
X: 5th metacarpal marker to 2nd
metacarpal marker
Left upper arm Frontal Z: elbow joint to shoulder joint Thorax– abdomen
-X: medial epicondyle to lateral epicondyle
Left forearm Frontal Z: wrist joint to elbow joint Left upper arm
-X: ulnar styloid marker to radial
styloid marker
Left hand Frontal Z: hand center point to wrist joint Left forearm
-X: 5th metacarpal marker to 2nd
metacarpal marker
Right thigh Frontal Z: knee joint to hip joint Pelvis
X: hip joint to lateral-thigh marker
Right shank Frontal Z: ankle joint to knee joint Right thigh
X: knee joint to lateral-shank marker
Right foot Sagittal Z: toe marker to heel marker Right shank
Y: hill marker to ankle joint
Left thigh Frontal Z: knee joint to hip joint Pelvis
-X: hip joint to lateral-thigh marker
Left shank Frontal Z: ankle joint to knee joint Left thigh
-X: knee joint to lateral-shank marker
Left foot Sagittal Z: toe marker to heel marker Left shank
Y: hill marker to ankle joint

Notes: In each frame, two axes are presented. The third axis is perpendicular to the anatomical plane formed by these
two axes. The first axis should be aligned accurately but the second axis can be recomputed from the first and third
axes. ASIS, right anterior-superior iliac spine; C7, cervical vertebra; SN, supra-sternal notch.

computed from the orientation matrices:

T i=j ¼ T i T0j ; ð2Þ

where j is the linked proximal segment, Ti/j is the relative orientation matrix to the proximal
segment, and T0 is the transpose of T. Relative orientation angles of the segments to their
respective linked proximal segments were computed from relative orientation matrices using
the mediolateral– anteroposterior –longitudinal (XYZ) rotation sequence:
2 3
C2 C3 S 1 S 2 C 3 þC 1 S 3 2C1 S 2 C3 þS 1 S 3
6 7
6 2C 2 S 3 2S 1 S 2 S 3 þC1 C 3 C 1 S 2 S 3 þS 1 C 3 7 ¼ T i=j ; ð3Þ
4 5
S1 2S 1 C 2 C1 C2
222 J. Kim et al.

where C and S are cos u and sin u, respectively, and u1, u2, and u3 are the relative orientation
angles. Angular velocities of the segments were then computed from the relative orientation
angles of the segments and their first time derivatives:
2 32 _ 3
C2 C3 S3 0 u1
6 7 6 7
vi ¼ vj þT0i 6 2C 2 S 3 C3 0 76 u_2 7; ð4Þ
4 54 5
S2 0 1 u_3

where v is the angular velocity. Angular momentums of the segments about the whole-body
CM were computed from positions, velocities, angular velocities, and inertial parameters of
the segments:

L i ¼ mi r~ i v i þðT0i I i T i Þvi ; ð5Þ

2 3
0 2zi yi
6 7
r~ i ¼ 6 z
4 i
0 2xi 7;
5 ð6Þ
2yi xi 0

2 3
I1 0 0
6 7
Ii ¼ 6
4
0 I2 0 7;
5 ð7Þ
0 0 I3

where L is the angular momentum, r is the relative position of the segmental CM to the
whole-body CM, r~ is the skew-symmetric form of r for cross-product operation, v is the
relative velocity of the segmental CM to the whole-body CM, I are the inertia tensor
described in the segmental reference frame, and I1, I2, and I3 are the principal MOI. The
segmental angular momentums were summed to yield angular momentums of the whole
body and body parts (lead [right] arm, trail [left] arm, head – trunk, gesture [right] leg, and
support [left] leg). Only the vertical angular momentum (z-component) was considered in
the analysis.
Five pirouette events were identified to facilitate data analysis (Figure 3): beginning of
double-stance (BDS; Figure 1C), beginning of angular momentum generation (BAG;
Figure 1D), gesture foot off (FO; Figure 1F), maximum vertical angular momentum
(MAM), and gesture foot contact (FC; Figure 1H). BDS was the instant dancer’s gesture
foot touched the rear force plate at the end of the preparatory motion. Dancers typically
started the double-stance phase (BDS to FO) with a slight counter (left) rotation of the
shoulder line and BAG was the instant dancer’s body initiated right turn at the end of
the counter rotation (Figure 3). Right turn of the body was initiated by an outward swing of
the lead arm (Figure 1D –E). In case the initial counter rotation was absent, BDS was used as
BAG.
The distance along the y-axis between the CMs of the feet at BAG was used as stance
depth (Figure 4A). The range of vertical position of the whole-body CM during the double-
stance phase (BDS to FO) was used as vertical motion range of the whole-body CM
(Figure 4A). The shoulder line vector (left to right shoulder joint center) was projected to the
ground to compute the projected angular position of the shoulder line with respect to the x-
 Angular momentum in pirouette en dehors 223

Figure 3. Exemplar ensemble-average patterns of the longitudinal angular momentum (Lz) of the whole body and
body parts (a female dancer; double-turn condition; n ¼ 3). Longitudinal angular momentum was normalized to
dancer’s mass2/height. A positive value means the body is turning right. BDS, beginning of double-stance; BAG,
beginning of angular momentum generation; FO, gesture foot off; MAM, maximum longitudinal angular
momentum; FC, gesture foot contact; WB, whole body; LA, lead (right) arm; TA, trail (left) arm; HT, head/trunk;
GL, gesture (right) leg; and SL, support (left) leg. The time axis was normalized to the BDS– FC time (in %).
Maximum longitudinal angular momentums of the whole body and body parts (black circles) during the angular
momentum generation phase (T; BAG to FO) were extracted and used in the analysis.

axis of the laboratory reference frame (Figure 4B). The third orientation angle of the upper
trunk (thorax– abdomen) relative to pelvis about the longitudinal axis (z-axis) was used as
the measure of trunk twist.
The following parameters were extracted from the computed kinematic and kinetic
variables for data analysis:

Figure 4. Kinematic parameters used: stance depth (d), vertical motion range of the whole-body center-of-mass
(CM; Dz), initial shoulder line angle (f1), and shoulder line angular displacement (Df). FO, gesture foot off; BAG,
beginning of angular momentum generation.
224 J. Kim et al.

. Duration of angular momentum generation [BAG – MAM time (in ms); Figure 3].
. Maximum vertical angular momentum of the whole body (Figure 3), normalized to
dancer’s ‘mass2/height’ (in m3/kg/s).
. Maximum vertical angular momentums of body parts (lead arm, trail arm, head – trunk,
gesture leg, and support leg) during the angular momentum generation phase
(Figure 3), normalized to the maximum whole-body angular momentum (in %).
. Average rate of angular momentum generation (maximum vertical angular momentum
divided by duration of angular momentum generation; whole body only), normalized to
dancer’s ‘mass2/height’ (in m3/kg/s2).
. Stance depth (Figure 4A), normalized to dancer’s height (in %).
. Vertical motion range of the whole-body CM (Figure 4A), normalized to dancer’s
height (in %).
. Initial shoulder line angle (at BAG) and shoulder line angular displacement (from BAG
to FO; Figure 4B) (in 8).
. Maximum trunk twist angle (in 8).

Statistical analysis
The values of the parameters obtained from the three repeated trials were averaged in each
dancer and each turn condition and the average values were used in the statistical analyses. A
one-way repeated-measure multivariate analysis of variance (MANOVA) was performed to
compare duration of angular momentum generation, maximum vertical angular momentum
of the whole body, rate of vertical angular momentum generation, stance depth, vertical CM
motion range, initial shoulder line angle, shoulder line angular displacement, and maximum
trunk twist angle among the turn conditions (single, double, and triple). A two-way (3 £ 5)
repeated-measure analysis of variance (ANOVA) was conducted to compare the maximum
angular momentum of the body part among turn conditions and body parts (lead arm, trail
arm, head –trunk, support leg, and gesture leg). Huynh –Feldt correction was used in case
the sphericity (and circularity) assumption was violated. Sidak correction was used for
multiple comparisons in the post hoc tests. In all statistical analyses, a was set to 0.05. All
statistical analyses were conducted using SPSS 15.0 for Windows.

Results
Figure 3 shows exemplar vertical angular momentum patterns of a female dancer (double-
turn; n ¼ 3). Peak vertical angular momentum of the whole body was observed slightly after
FO. The arms reached their respective peak values during the angular momentum
generation phase but the other body parts reached their peak values during the single-stance
phase (FO to FC). For head – trunk and legs, vertical angular momentum values at MAM
were used as the maxima.
A significant ‘turn’ effect was observed in the one-way MANOVA (Wilks’ l ¼ 0.014,
F16,18 ¼ 8.505, p , 0.001, h2p ¼ 0.883). Post hoc analysis revealed that the ‘turn’ effect
mainly came from five parameters: maximum vertical angular momentum of the whole body
(triple . double . single), vertical angular momentum generation rate (triple . double .
single), vertical CM motion range (single . triple), shoulder line angular displacement
(triple/double . single), and maximum trunk twist (triple/double . single) (Table III). The
vertical CM motion range showed a decreasing trend as the number of turns increased.
Comparison of the normalized maximum vertical angular momentum among the turn
conditions and body parts revealed significant ‘turn’ effect (F1.821 ¼ 4.169, p ¼ 0.040,
 Angular momentum in pirouette en dehors 225

Table III. Summary of select kinematic and kinetic parameters (n ¼ 9; M ^ SD)

Turn conditions

Variable Single Double Triple

Maximum AM (£ 10-3 m3/kg/s) 4.8 ^ 0.6 6.9 ^ 0.8* (þ 45.2%) 7.7 ^ 1.0*$ (þ 62.4%)
Duration of AM generation (ms) 555 ^ 124 578 ^ 103 (þ 5.4%) 531 ^ 72 (-0.9%)
AM generation rate (£ 10-3 m3/kg/s2) 9.2 ^ 3.2 12.5 ^ 3.4* (þ 39.1%) 14.7 ^ 2.4*$ (þ 69.0%)
Stance depth (%) 22.8 ^ 4.3 22.7 ^ 4.2 (-0.2%) 22.8 ^ 4.4 (þ 0.4%)
Vertical CM motion range (%) 7.8 ^ 1.1 6.9 ^ 1.5 (-12.3%) 6.1 ^ 1.6* (-20.8%)
Initial shoulder line angle (8) 26.4 ^ 4.9 26.7 ^ 5.0 (þ 2.2%) 28.1 ^ 5.7 (þ 7.7%)
Shoulder line rotation range (8) 101.7 ^ 14.9 141.5 ^ 12.8* (þ 40.5%) 151.4 ^ 16.4* (þ 50.5%)
Maximum trunk twist (8) 28.9 ^ 5.3 31.6 ^ 5.8* (þ 9.8%) 32.9 ^ 3.7* (þ 15.4%)

Notes: Values in parentheses are net changes from the single-turn condition. AM, longitudinal angular momentum;
CM, center-of-mass. *Significantly different from the single-turn condition (p , 0.05). $Significantly different from
the double-turn condition (p , 0.05).

h2p ¼ 0.343), ‘body’ effect (F3.716 ¼ 209.795, p , 0.001, h2p ¼ 0.963), and interaction
(F3.529 ¼ 17.859, p , 0.001, h2p ¼ 0.691). The trail arm showed the largest maximum
vertical angular momentum during the angular momentum generation phase, followed by
the lead arm and head –trunk. The legs were characterized by the smallest maximum vertical
angular momentum (Table IV). Among the body parts, lead arm (single . double) and trail
arm (single . double . triple) showed decreasing trends as the number of turns increased,
whereas head –trunk (double/triple . single), right leg (double . single), and left leg (triple/
double . single) revealed increasing trends (Table IV).

Discussion and implications

Skilled dancers in this study generated significantly more vertical angular momentum as the
demand (number of turns) increased (þ 45.2% from the single- to double-turn condition and
þ 62.4% from the single- to triple-turn condition; Table III). This was accomplished mainly
by increasing the rate of angular momentum generation (þ 39.1% and þ 69.0%, respectively)
while maintaining the duration of angular momentum generation relatively unchanged

Table IV. Maximum normalized angular momentums of body parts (n ¼ 9; M ^ SD; in %)

Turn conditions

Variable Single Double Triple

Lead arm (LA) 29.8 ^ 4.8 27.0 ^ 3.1* 26.7 ^ 3.0

Trail arm (TA) 43.5 ^ 2.1 39.6 ^ 1.9* 37.9 ^ 2.1*$
Head–trunk (HT) 15.7 ^ 2.8 17.7 ^ 2.6* 17.5 ^ 2.2*
Gesture leg (GL) 12.0 ^ 1.2 13.6 ^ 2.3* 13.8 ^ 2.2
Support leg (SL) 11.2 ^ 1.2 12.6 ^ 1.2* 13.3 ^ 2.1*
Post hoc TA . LA . HT . SL TA . LA . HT . SL TA . LA . HT . SL
TA . LA . GL TA . LA . GL TA . LA . GL

Note: Maximum vertical angular momentums of the body parts were normalized to the maximum vertical angular
momentum of the whole body.
*Significantly different from the single-turn condition (p , 0.05).
$
Significantly different from the double-turn condition (p , 0.05).
226 J. Kim et al.

(Table III). Generation of a larger angular momentum is necessary to offset the friction
between the stance foot and the floor as well as the distance of the arms en avant and the
gesture leg in retiré from the axis of rotation. The stance depth changed by less than 1%
(Table III), suggesting that the dancers increased the vertical angular momentum by
increasing the ground reaction torque about the vertical axis without the benefit of increased
moment arms in the double- and triple-turn conditions.
Among the body rotation parameters, initial shoulder line angle did not show any
significant changes while shoulder line rotation range and maximum trunk twist revealed
significant differences between the single-turn condition and the double-/triple-turn
conditions (Table III). Moreover, the actual increases in maximum trunk twist (þ 2.78 and
þ 4.08 from single to double and single to triple, respectively) were substantially smaller than
those in shoulder line rotation range (þ 39.88 and þ 49.78, respectively; Table III). This
suggests that the initial position of the shoulder line and the initial trunk twist at BAG were
not much different among the turn conditions. Angular momentum generation is typically
initiated with a twisting (wind-up) motion of the upper trunk relative to pelvis and the en bloc
motion of the trunk (as one unit) starts when the shoulder line returns to a parallel alignment
with the hip line during the clockwise rotation. Similar initial shoulder line positions and
marginal changes in the maximum trunk twist can be translated to marginal changes in wind-
up motion range, thus meaning substantial changes in en bloc rotation range of the trunk.
Interestingly, the vertical CM motion range revealed a decreasing trend as the number of
turns increased with a significant difference between the single- and triple-turn conditions
(Table III). In other words, as the number of turns increased, skilled dancers tended to
suppress the down-and-up motion of the body while promoting rotation of the trunk
(shoulder line) and arms in order to generate additional angular momentum. With increased
shoulder line rotation range and maximum trunk twist, the dancer can create an increased
muscular counter-tension in the body which facilitates the rotation of the body with respect
to the ground.
Among the body parts, the trail (left) arm showed the largest normalized peak vertical
angular momentum (37.9 –43.5% of the whole-body peak vertical angular momentum)
during the angular momentum generation phase, followed by the lead (right) arm (26.7 –
29.8%; Table IV). Arm rotations are essentially sum of the shoulder line rotation and the
relative rotations of the arms to shoulder line. In the middle of the angular momentum
generation phase, the direction of relative rotation of the lead arm to shoulder line about the
vertical axis reverses in order to prepare for the upcoming en avant position, while the trail
arm continues a relative rotation in the direction of turn throughout the entire angular
momentum generation phase. As a result, the lead arm reaches its peak vertical angular
momentum earlier than the trail arm and the peak value is also smaller (Figure 3).
Traditionally, emphasis has been placed on the timing of the lead arm in pirouette teaching
but it is important for the practitioners to understand the role of the trail arm in angular
momentum generation during the double-stance phase (Warren, 1989).
The arms start losing vertical angular momentum after reaching their respective peak
values during the angular momentum generation phase while the other body parts showing
continuous increases throughout the same phase (Figure 3), which suggests an angular
momentum transfer from the arms to the rest of the body. It appears that this transfer of
angular momentum is mainly caused by the transition from ‘wind-up’ to en bloc motion of the
trunk (Golomer et al., 2009). The upper trunk angular velocity (and those of the arms)
decreases during the transition from wind-up to en bloc motion of the trunk as upper trunk
starts moving with the pelvis as one unit. Arm motions toward the body in the en avant arm
position may also facilitate the loss of angular momentum in the arms during the en bloc
 Angular momentum in pirouette en dehors 227

motion of the trunk. Bringing the arms closer to the body decreases the MOI of the arms,
which, in turn, decreases the angular momentums of the arms. In summary, the arms (the
trail arm in particular) are the main contributors to the whole-body angular momentum and
an effective angular momentum generation strategy in pirouette en dehors must focus on the
coordination of the arm motions and trunk, and trail arm motion in particular (Warren,
1989).
In this study, the head segment was combined with the trunk in assessing contributions of
the segments to the whole-body angular momentum. Although spotting introduces a rapid
relative motion of the head to the trunk, the maximum angular momentums of the head were
only 1.7%, 2.7%, and 2.7% of that of the whole body for the single-, double-, and triple-turn
conditions, respectively. This suggests that spotting has a negligible effect in generating
angular momentum during the double-support phase of a pirouette.
The proportions of the arm angular momentums to whole-body angular momentum
decreased as the number of turns increased (Table IV), which was mainly because of a more
dominant increase in peak whole-body angular momentum than that in peak arm angular
momentums. As the number of turns increases, the proportion of en bloc motion of the trunk
increases. A wind-up trunk motion during the initial double-stance phase puts more
emphasis on the upper trunk and shoulder girdle/arm rotations, whereas an en bloc motion of
the whole trunk facilitates rotation of the head/trunk and legs as well. The decreased
proportions of the arm angular momentums and increased proportions of the head/trunk
and leg angular momentums with more turns thus are normal.
Whereas the dancer is generally advised not to accentuate the shoulder or arm movement
in the preparation for the turn, there is an evident mechanical advantage of maximizing
shoulder and arm motions in generating large angular momentum during the double-
support phase of a pirouette en dehors (Herstens, 1989). Further studies may look at the
optimum movement magnitudes and timings for the motions of individual body parts (upper
trunk and arms in particular) as well as comparing stylistic differences in ballet styles which
‘do’ or ‘do not’ favor shoulder and torso rotations during the pirouette preparation (Biringen,
2010; Denardi & Corrêa, 2013; Laws, 2002; Lott & Laws, 2012; Warren, 1989).
There can be two main reasons for a dancer’s inability to successfully perform a pirouette
with multiple turns: (1) insufficient angular momentum generated during the double-stance
phase and/or (2) inability to maintain balance during the single-stance phase. Further studies
involving successful and unsuccessful trials of skilled dancers as well as those involving skilled
and unskilled dancers may shed light on this particular issue. Another important area that
deserves further investigation is the transition from double- to single-stance, in which the
CM of the dancer’s body must be shifted toward the support foot and raised up without
losing balance to assume a stable pointe (or demi pointe) position.
The preparation for the pirouette en dehors in this study was limited to a relevé passé
preparation, rather than a tendu/demi rond de jambe to fourth position, or a turn from fifth or
second position. In addition, the end of the preparation movement was demi plié in fourth
position, rather than an allongé position. These stylistic variations in the preparation were not
considered in this study and the findings of the study may not be generalized to pirouette styles
with different preparation scenarios. Moreover, ballet technique shoes instead of pointe shoes
were used in this study. Due to different friction coefficients, shoe conditions can affect the
rate of angular momentum generation during the double-stance phase and loss during the
single-stance phase. Therefore, caution must be exercised in generalizing the findings of this
study in terms of the footwear. Further investigations focusing on the effect of shoe
conditions and various preparation scenarios on angular momentum generation profile are
warranted.
228 J. Kim et al.

One limitation of this study was the relatively small sample size (n ¼ 9) used including one
male dancer. While male dancers generally are able to perform more turns than female
dancers, it was assumed in this study that they would respond to the three turn conditions
(single, double, and triple) similarly. Most of the parameters (maximum angular momentum,
angular momentum generation rate, stance depth, vertical CM motion range, and maximum
angular momentums of the body parts) were normalized to eliminate the effects of body size
and the values of the male dancer were well within the value ranges of the female participants in
most cases. The male dancer showed on average two extreme values (the largest or the smallest
among the participants) out of 26 possible with the average expectation being 2.9 per
participant (13 parameters £ 2 extreme values per parameter / 9 participants). Removal of the
male dancer from statistical analyses revealed only minor changes in the univariate ANOVA
results, which could also have occurred with removal of any female participant. Inclusion of
the male dancer in the data analysis (n ¼ 9) enabled a one-way repeated-measures MANOVA
with all eight angular momentum and motion strategy parameters (Table III). Nevertheless,
larger scale investigations of gender effects on angular momentum generation strategies are
warranted.

Conclusion
The purpose of this study was to investigate the vertical angular momentum generation
strategies used by skilled ballet dancers in pirouette en dehors. For this, three turn conditions
(single, double, and triple turns) were systematically used and select linear and angular
motion parameters along with the peak angular momentums of the body parts were
compared among the turn conditions.
Skilled dancers generate larger vertical angular momentum as the number of turns
increases by predominantly increasing the rate of momentum generation. To generate
additional angular momentum, skilled dancers use a strategy which promotes the rotation of
the upper trunk and arms without varying the time to generate the turning motion. First, the
dancers use a twisting (wind-up) motion of the trunk relative to the pelvis, and an en bloc
motion of the trunk after the turn is initiated. Second, the action of the arms is important
with the trail (left) arm contributing to angular momentum generation the most. The trail
arm rotates in the direction of the turn so it can contribute more to the generation of angular
momentum during the double-stance phase, the preparation phase of the turn. An effective
angular momentum strategy must, therefore, focus on the arm motions and shoulder girdle
motions in pirouette en dehors.

References
Bell, A. L., Pedersen, D. R., & Brand, R. A. (1990). A comparison of the accuracy of several hip center location
prediction methods. Journal of Biomechanics, 23, 617–621.
Biringen, E. (2010). Analysis of pirouette execution for improved performance. Medical Problems of Performing
Artists, 25, 136– 137.
de Leva, P. (1996). Adjustments to Zatsiorsky-Seluyanov’s segment inertia parameters. Journal of Biomechanics, 29,
1223–1230.
Denardi, R. A., & Corrêa, U. C. (2013). Effects of instructional focus on learning a classical ballet movement, the
pirouette. Journal of Dance Medicine & Science, 17, 18–23.
Golomer, E., Toussaint, Y., Bouillette, A., & Keller, J. (2009). Spontaneous whole body rotations and classical dance
expertise: How shoulder– hip coordination influences supporting leg displacements. Journal of Electromyography
and Kinesiology, 19, 314–321.
Herstens, M. (1989). Ballet technique: Principles for the horizontal floor. Saratoga, CA: R & E Publishers.
 Angular momentum in pirouette en dehors 229

Kwon, Y. -H. (2008). Measurement for deriving kinematic parameters: Numerical methods. In Y. Hong & R. M.
Bartlett (Eds.), Routledge handbook of biomechanics and human movement science (pp. 156–181). Oxfordshire:
Routledge.
Laws, K. (1984). The physics of dance. New York, NY: Schirmer Books.
Laws, K. (1998). Momentum transfer in dance movement. Medical Problems of Performing Artists, 13, 136–145.
Laws, K. (2002). Physics and the art of dance. New York, NY: Oxford University Press.
Lott, M. B., & Laws, K. L. (2012). The physics of toppling and regaining balance during a pirouette. Journal of
Dance Medicine & Science, 16, 167–174.
MacKinnon, C. D., & Winter, D. A. (1993). Control of whole body balance in the frontal plane during human
walking. Journal of Biomechanics, 26, 633–644.
Sugano, A. (2010). Applications of dance science to the training of dancers. In Proceedings of the 20th annual
international dance medicine and science meeting. Birmingham: International Association for Dance Medicine and
Science.
Sugano, A., & Laws, K. (2002). Physical analysis as a foundation for pirouette training. Medical Problems of
Performing Artists, 18, 29– 32.
Warren, G. W. (1989). Classical ballet technique. Gainesville: University Press of Florida.
Copyright of Sports Biomechanics is the property of Routledge and its content may not be
copied or emailed to multiple sites or posted to a listserv without the copyright holder's
express written permission. However, users may print, download, or email articles for
individual use.