A three-dimensional biomechanical

analysis of sumo and conventional

style deadlifts
RAFAEL F. ESCAMILLA, ANTHONY C. FRANCISCO, GLENN S. FLEISIG, STEVEN W. BARRENTINE,
CHRISTIAN M. WELCH, ANDREW V. KAYES, KEVIN P. SPEER, and JAMES R. ANDREWS
Michael W. Krzyzewski Human Performance Laboratory, Division of Orthopaedic Surgery, Duke University Medical
Center, Durham, NC 27710; American Sports Medicine Institute, Birmingham, AL 35205; and Human Performance
Technologies, Inc., Jupiter, FL 33477

ABSTRACT
ESCAMILLA, R. F., A. C. FRANCISCO, G. S. FLEISIG, S. W. BARRENTINE, C. M. WELCH, A. V. KAYES, K. P. SPEER, and
J. R. ANDREWS. A three-dimensional biomechanical analysis of sumo and conventional style deadlifts. Med. Sci. Sports Exerc., Vol.
32, No. 7, pp. 1265–1275, 2000. Purpose: Strength athletes often employ the deadlift in their training or rehabilitation regimens. The
purpose of this study was to quantify kinematic and kinetic parameters by employing a three-dimensional analysis during sumo and
conventional style deadlifts. Methods: Two 60-Hz video cameras recorded 12 sumo and 12 conventional style lifters during a national
powerlifting championship. Parameters were quantified at barbell liftoff (LO), at the instant the barbell passed the knees (KP), and at
lift completion. Unpaired t-tests (P ⬍ 0.05) were used to compare all parameters. Results: At LO and KP, thigh position was 11–16°
more horizontal for the sumo group, whereas the knees and hips extended approximately 12° more for the conventional group. The
sumo group had 5–10° greater vertical trunk and thigh positions, employed a wider stance (70 ⫾ 11 cm vs 32 ⫾ 8 cm), turned their
feet out more (42 ⫾ 8° vs 14 ⫾ 6°), and gripped the bar with their hands closer together (47 ⫾ 4 cm vs 55 ⫾ 10 cm). Vertical bar
distance, mechanical work, and predicted energy expenditure were approximately 25– 40% greater in the conventional group. Hip
extensor, knee extensor, and ankle dorsiflexor moments were generated for the sumo group, whereas hip extensor, knee extensor, knee
flexor, and ankle plantar flexor moments were generated for the conventional group. Ankle and knee moments and moment arms were
significantly different between the sumo and conventional groups, whereas hip moments and moments arms did not show any
significantly differences. Three-dimensional calculations were more accurate and significantly different than two-dimensional calcu-
lations, especially for the sumo deadlift. Conclusions: Biomechanical differences between sumo and conventional deadlifts result from
technique variations between these exercises. Understanding these differences will aid the strength coach or rehabilitation specialist in
determining which deadlift style an athlete or patient should employ. Key Words: POWERLIFTING, WEIGHTLIFTING, JOINT
MOMENTS, JOINT MOMENT ARMS, JOINT ANGLES, SEGMENT ANGLES, KINEMATICS, KINETICS, MECHANICAL
WORK

T
he deadlift, which measures overall body strength, is
the last of three lifts in powerlifting competition. The
starting position for the deadlift is with the lifter in a
squat position, arms straight and pointing down, and an
alternating hand grip used to hold a bar positioned in front
of the lifter’s feet. According to the American Drug Free
Powerlifting Association (ADFPA) rules at the time of this
study, the barbell is lifted upward in a continuous motion
until the lifter is standing erect with knees locked and the
shoulders thrust back. Causes for disqualification included
failure to wait for the referee’s “down” signal at the com-
pletion of the lift, any stopping or downward movement of
the bar once the bar leaves the lifting platform, failure to
0195-9131/00/3207-1265/0
MEDICINE & SCIENCE IN SPORTS & EXERCISE®
Copyright © 2000 by the American College of Sports Medicine
Submitted for publication April 1999.
Accepted for publication October 1999.
1265
stand erect with locked knees and shoulders thrust back, and
any “hitching,” bouncing, or resting of the bar against the
thighs during the lift. All deadlift trials analyzed in the
current study were in accordance with these rules.
Strength athletes, such as American football players, of-
ten employ the barbell deadlift in their training or rehabil-
itation regimens. These athletes use the deadlift to enhance
hip, thigh, and back strength. The deadlift is performed
using either a conventional or sumo style. The primary
differences between these two styles are that the feet are
positioned further apart and turned out in the sumo style,
and the arms are positioned inside the knees for the sumo
style and outside the knees for the conventional style. Stance
width, foot angle, and hand width differences between these
two styles have not yet been quantified. Although both
deadlift styles are used in training, the efficacy of one style
over another is unclear.
Because the deadlift is considered a closed kinetic chain
exercise (23), it can also be employed in knee rehabilitation
programs, such as after anterior cruciate ligament (ACL)
reconstruction. Numerous studies have already shown that
the squat is an effective exercise during ACL rehabilitation
(11,15,20,22,25,26,29). Because the deadlift is performed in
a similar manner as the squat, it is hypothesized that the
deadlift may provide similar benefits during ACL rehabili-
tation. The moderate to high hamstring activity that has been
reported during the deadlift (28) may help protect the ACL
during knee rehabilitation. However, which deadlift style
would be most effective in knee rehabilitation has not yet
been established.
There are seven known studies that have examined bio-
mechanical variables during the barbell deadlift (2,4,5,8 –
10,17). Three studies examined lumbar spinal loads (4,5,8),
two studies investigated the effects of intra-abdominal and
intra-thoracic pressures (9,10), one study quantified joint
and segmental angles (17), and the remaining study calcu-
lated joint angles and joint moments (2). However, only two
studies have compared kinematic or kinetic parameters be-
tween sumo and conventional deadlifts (5,17). McGuigan
and Wilson (17) performed a kinematic analysis using male
lifers from two regional powerlifting championships. The
only significant differences they observed were that the
sumo group had a more upright trunk and less hip flexion at
liftoff, and the shank range of motion was greatest in the
sumo group. Cholewicki et al. (5) quantified lumbar loads
and hip and knee moments between the sumo and conven-
tional deadlifts during a national powerlifting champion-
ship. They found significantly greater L4 –L5 shear forces
and moments in the conventional group, whereas hip and
knee moments were not significantly different between the
two deadlift styles. One limitation to all previous deadlift
studies was that two-dimensional (2-D) analyses (i.e., one
camera employed to record a sagittal view of the lifter) were Figure 1—Sumo (top) and conventional (bottom) style deadlifts.
conducted in quantifying kinematic and kinetic parameters.
Although trunk movements through spinal and hip flexion
and extension occur primarily in the sagittal plane, flexion MATERIALS AND METHODS
and extension movements at the ankle and knee occur in the Subjects. Twenty-four male powerlifters served as sub-
sagittal plane only if the feet are positioned in that plane jects. Twelve of the subjects performed the conventional
(i.e., pointing straight ahead). This is because the ankles and deadlift, whereas the remaining 12 subjects performed the
knees primarily function as hinge joints and thus move in sumo deadlift (Fig. 1). All subjects wore a one-piece lifting
the direction the feet point. Therefore, the lower extremities suit. Mean age, body mass, and body height were 47.4 ⫾ 7.3
will move out of a sagittal plane as the feet turn outward. yr, 71.6 ⫾ 10.8 kg, and 172 ⫾ 8 cm, respectively, for the
These will cause erroneous measurements of lower extrem- sumo group, and 46.4 ⫾ 6.1 yr, 76.8 ⫾ 22.7 kg, and 170 ⫾
ity joint and segment angles, ankle and knee moments, and 8 cm, respectively, for the conventional group. All subjects
moment arms. These errors may be minimal during the participated in a national powerlifting masters’ champion-
conventional deadlift, because the feet point either straight ship that was sanctioned by the ADFPA. To participate in
ahead or are slightly turned out, but considerable errors can masters’ level powerlifting competition, all lifters had to be
occur during the sumo deadlift, because the feet turn out to at least 40 yr old. All subjects signed a human consent form
a greater degree. Therefore, it was the purpose of this study giving their approval to be videotaped and participate in this
to compare joint and segment angles and ankle, knee, and study.
hip moments and moment arms using three-dimensional Data collection. Two synchronized Sony HVM 200
(3-D) analyses during sumo and conventional deadlifts. video cameras were used to collect 60-Hz video data. One
Mechanical work and predicted energy expenditure was also camera faced the subject’s left side whereas the other cam-
quantified. A secondary purpose was to compare kinematic era faced the subject’s right side, with each camera’s optical
and kinetic calculations between 2-D and 3-D analyses for axis forming a 45° angle to the sagittal plane of the lifter.
both sumo and conventional deadlifts. The cameras were positioned approximately14 m apart and
1266 Official Journal of the American College of Sports Medicine http://www.msse.org
faced perpendicular to each other, with each camera approx-
imately10 m from the subject. To minimize the effects of
digitizing error, the cameras were positioned so that the
lifter-barbell system was as large as possible within the
viewing area of the cameras.
For each camera view, videotaping began approximately
1 min before a subject starting his lift and ended approxi-
mately 1 min after the completion of their lift. Just before a
subject beginning their lift, an external light source was
activated in both camera views to help match video frames
when viewing the two videotapes. Before and just after the
subjects were videotaped, a 2 ⫻ 1.5 ⫻ 1 m 3-D calibration
frame (Peak Performance Technologies, Inc., Englewood,
CO), surveyed with a measurement tolerance of 0.5 cm, was
positioned and videotaped in the same volume occupied by
the lifter-barbell system. The calibration frame was com-
prised of 24 spherical balls of known spatial coordinates,
with the X- and Z-axes positioned parallel to the ground,
and the Y-axis pointing vertical.
Data analysis. In powerlifting competition, a lifter is
given three attempts during the deadlift to maximize the
amount of weight they can lift. A lifter’s first attempt is
usually submaximal, whereas their second and third at-
tempts are near the maximal weight they are capable of
lifting. Therefore, only second and third attempts that were
successfully completed (i.e., ruled a “good lift” by a panel
of three judges) were analyzed. Seventeen of the 24 lifts
analyzed were third attempts. The seven second-attempt lifts
were used because the third attempts were unsuccessful due
to the lifter attempting a weight that was beyond their one
repetition maximum (1 RM). Therefore, it was thought that
all lifts analyzed were very near each lifter’s 1 RM.
Three events were defined during the deadlift. The first
event was barbell liftoff (LO), which was defined as the first
picture in which the barbell disks on both sides of the bar
were no longer in contact with the lifting platform. Because
both sides of the bar typically remained symmetrical
throughout the lift, the left and right side barbell disks left
the lifting platform at approximately the same time. There-
fore, at LO the lifter was supporting the entire barbell load.
The next event was at the instant the bar passed the knees
(KP), which was defined as the first picture when the ver-
tical position of the bar was higher than the vertical position
of the knees. The last event was lift completion (LC), which
occurred when the lifter was in an upright position with the
knees and hips fully extended and the shoulders pulled back.
At this time, the head judge gave the “down” command,
signaling the end of the lift. Total lift time was defined from
LO to LC.
A 3-D video system (Peak Performance Technologies,
Inc., Englewood, CO) was used to manually digitize data for
all 24 subjects. A 15-point spatial model was created, com-
prised of the top of the head and centers of the left and right
mid-toes, ankles, knees, hips, shoulders, hands, and end of
bar. All points were seen in each camera view. Each of these
15 points was digitized in every video field (60 Hz), which
was adequate due to the slow movement of the lift (2,5,14).
To minimize manual digitizing error, each point was digi-
BIOMECHANICAL ANALYSIS OF THE DEADLIFT
tized twice and averaged. Digitizing began 15 video fields
(0.25 s) before LO and ended 15 video fields after LC.
A fourth-order, zero-lag Butterworth digital filter was
used to smooth the raw data with a cutoff frequency of 5 Hz.
A cutoff frequency between 3 and 5 Hz has been demon-
strated to be adequate during lifting 1 RM loads involving
slow movements (7,17). By using the direct linear transfor-
mation method (24,27), 3-D coordinate data were derived
from the 2-D digitized images from each camera view. An
average resultant mean square calibration error of 0.8 cm
produced an average volume percent error of 0.239.
The origin of the 3-D orthogonal axis system was first
translated to the right ankle joint and rotated so that the
positive X-axis pointed to the left ankle joint, the positive
Z-axis pointed anteriorly in the direction the lifter was
facing, and the Y-axis pointed in the vertical direction. The
vertical positions of the digitized left and right ankles were
within 1 cm of each other. This axes system was initially
used to calculate all joint moments and moment arms and
joint and segment angles. Muscle moment arms were not
quantified in this study. Because hip flexion and extension
during sumo and conventional deadlifts occur primarily in
the Y-Z sagittal plane about the X-axis, hip moments were
calculated about the X-axis and hip moment arms were
calculated in the Z-axis direction. Ankle and knee moment
arms were also calculated in the Z-axis direction, which
equates to a 2-D analysis using one camera to record a
sagittal view of the lifter. These 2-D data were compared
with 3-D data from the 3-D analysis. Because during the
sumo deadlift the feet were turned out approximately 45°,
ankle and knee flexion and extension occurred in a plane
midway between the Y-Z sagittal plane and X-Y frontal
plane established above. Therefore, erroneous moment arm
measurements would occur if the above axes system were
used for the 3-D analysis, because ankle and knee move-
ments do not occur in the sagittal plane during the sumo
deadlift. Because the feet are only slightly turned out during
the conventional deadlift, relatively small errors would re-
sult in moment arm calculations compared with the sumo
deadlift. Therefore, the axes system was translated to each
ankle joint center and rotated so that the positive Z-axis
pointed in the direction of the mid-toes, the Y-axis pointed
vertical, and the X-axis was orthogonal to the Y- and Z-
axes. Hence, for both sides of the body, ankle and knee
moments were calculated about the X-axis, and ankle and
knee moment arms were calculated in the Z-axis direction.
Linear and angular displacements and velocities were
calculated for both the left and right sides of the body and
then averaged. Relative knee and hip angles and absolute
trunk, thigh, and shank angles were defined in accordance
with previous studies (2,17). Trunk, thigh, and shank angles
were measured relative to the X-Z horizontal plane (i.e.,
from a right horizontal relative to a sagittal view of the
lifter’s right side). Knee angles were measured relative to
the thigh and leg segments. Although the hip angle is actu-
ally formed between the pelvis and the femur, this measure-
ment cannot accurately be determined without external
markers affixed to these segments. Therefore, the hip angle
Medicine & Science in Sports & Exercise姞 1267
was defined as the relative angle between the trunk (hip to
shoulder segment) and thigh (hip to knee segment). As long
as the trunk is rigid and straight, this relative angle approx-
imates the true hip angle. However, during the deadlift, the
trunk does not remain rigid and straight, because spinal
flexion causes the back to round to some extent, especially
when maximum weight is used. As the spine flexes during
the deadlift, the shoulders drop downward, causing the hip
angle measurement to decrease and underestimate its true
value. To compare joint and segment angle differences
between a 3-D and 2-D analysis, hip and knee relative
angles and thigh and shank segment angles from the 3-D
analysis were projected onto a 2-D sagittal plane. Foot angle
was defined as the angle formed between the foot segment
and the Y-Z sagittal plane. Stance width was defined as the
linear distance between the left and right ankle centers,
whereas hand width was defined as the linear distance
between the left and right hand centers.
During 1 RM lifting, studies have shown (7,13) that the
barbell initially accelerates at LO to a first peak velocity
(acceleration phase), then decelerates to a minimum velocity
(sticking region), accelerates again to a second peak velocity
(maximum strength region), and finally decelerates until LC
(deceleration phase). These lifting phases and regions were
also calculated in the current study. The Schwartz score,
which is used in powerlifting competition to determine the
overall best lifter for an event, was used to normalize the
barbell load to each lifter’s body mass and to compare
relative loads lifted between sumo and conventional groups.
A complete list of Schwartz coefficients for any given body
weight can be obtained in official powerlifting rule books.
Because segment and barbell accelerations are very small
while lifting maximum or near maximum loads, joint mo-
ments can accurately be calculated using quasi-static models
(12,14,18,19,21). Lander et al. (14) found that joint mo-
ments varied less than 1% between quasi-static and dynamic
analyzes during the squat exercise with near maximum
loads. Left and right hip, knee, and ankle moments and
moment arms were calculated at LO, KP, and LC and then
averaged. Body segment center of masses and weights were
calculated by using appropriate anthropometric data (6) and
each lifter’s known mass. Joint moments and moment arms
were calculated relative to both barbell weight and system
weight. The system weight used to calculate joint moments
was the sum of the barbell weight and the weight of body
segments above the joint in which the moments were cal-
culated. Therefore, the weights of the trunk, neck, head, and
upper extremities were used to calculate system moments at
the hip; the weights of the trunk, neck, head, upper extrem-
ities, and thighs were used to calculate system moments at
the knee; and the entire body weight (minus the feet) was
used to calculate system moments at the ankle. The geo-
metric center of the barbell represented the center of mass of
the barbell (COMbar). X, Y, and Z position coordinates were
calculated for both COMbar and the center of mass of the
system (COMsystem). Ankle moment arms (MAankle) were
calculated as the distance in the Z-axis direction from the
ankle joints to COMsystem or COMbar. Ankle moments were
1268 Official Journal of the American College of Sports Medicine
the product of MAankle and system/barbell weight. Knee
moment arms (MAknee) were calculated as the distance in
the Z-axis direction from the knee joints to COMsystem or
COMbar. Knee moments were the product of MAknee and
system/barbell weight. Hip moment arms (MAhip) were
calculated as the distance in the Z-axis direction from the
hip joints to COMsystem or COMbar. Hip moments were the
product of MAhip and system/barbell weight.
Because bar motion primarily occurred in the vertical
direction, vertical bar displacement was calculated from LO
to LC and normalized by body height. Mechanical work,
which was calculated relative to both barbell weight and
system weight, was the product of system or barbell weight
and total vertical displacement of COMbar or COMsystem.
The system weight used to calculate mechanical work was
the sum of barbell weight and body weight. Independent
t-tests (P ⬍ 0.05) were used to compare kinematic and
kinetic parameters between sumo and conventional deadlift
groups, while paired t-tests were used to compare 2-D and
3-D comparisons.
RESULTS
Of the 110 powerlifters observed during the national
powerlifting competition described in the current study,
70% used the conventional style, whereas 30% used the
sumo style. However, of the 54 lifters in the heavier weight
classes (90 –125⫹ kg), 85% used the conventional style and
only 15% used the sumo style. In contrast, of the 56 lifters
in the lighter weight classes (52– 82 kg), only 55% used the
conventional style, whereas 45% used the sumo style.
Therefore, it appears that heavier lifters more commonly use
the conventional style, whereas lighter lifters are more
evenly split between the two lifting styles.
Comparisons of joint and segment angles between sumo
and conventional deadlifts are shown in Table 1. Compared
with the conventional group, at LO the sumo group main-
tained a more upright trunk, positioned the thigh closer to
the horizontal, and positioned the shank closer to the verti-
cal. At KP and at minimal bar velocity, the sumo group had
greater hip and knee flexion (i.e., smaller relative hip and
knee angles), whereas the conventional group positioned the
shank closer to the vertical and the thigh closer to the
horizontal. From LO to KP, the conventional group ex-
tended their hips, knees, and shank through a greater range
of motion than the sumo group. Throughout the lift, the feet
were turned outward 10 –15° for the conventional group and
40 – 45° for the sumo group. Table 2 shows joint and seg-
ment differences between a 2-D and 3-D analysis. For the
sumo deadlift, a 2-D analysis was significantly different
than a 3-D analysis for all measurements. In contrast, for the
conventional deadlift a 2-D analysis was not significantly
different from a 3-D analysis for all measurements except
shank angle at KP.
From Table 3, the sumo group employed a foot stance
that was 2–3 times wider than the conventional group,
whereas the conventional group had a 17% greater hand
width than the sumo group. There were no significant dif-
http://www.msse.org
TABLE 1. Joint and segment angles (mean ⫾ SD) between sumo and conventional deadlifts.
Sumo (mean ⴞ SD) Conventional (mean ⴞ SD)
McGuigan and McGuigan and Brown and
Current Wilson (17) Current Wilson (17) Abani (2)
Liftoff (LO)
Hip (°) 77 ⫾ 7 77 ⫾ 9** 72 ⫾ 12 67 ⫾ 5** 69 ⫾ 5
Knee (°) 126 ⫾ 8 127 ⫾ 8 124 ⫾ 9 120 ⫾ 10 123 ⫾ 6
Trunk (°) 33 ⫾ 11* 25 ⫾ 8* 24 ⫾ 10* 17 ⫾ 7* 23 ⫾ 6
Thigh (°) 145 ⫾ 9** 136 ⫾ 6 134 ⫾ 7** 137 ⫾ 5 134 ⫾ 4
Shank (°) 80 ⫾ 4** 63 ⫾ 11 76 ⫾ 5** 59 ⫾ 5 76 ⫾ 3
Knee passing (KP)
Hip (°) 106 ⫾ 8* 128† 114 ⫾ 11* 122† 134 ⫾ 18
Knee (°) 153 ⫾ 8* 168† 161 ⫾ 8* 165† 165 ⫾ 6
Trunk (°) 46 ⫾ 7 48† 43 ⫾ 7 43† 59 ⫾ 14
Thigh (°) 126 ⫾ 9** 106† 110 ⫾ 5** 107† 105 ⫾ 4
Shank (°) 79 ⫾ 5** 70† 84 ⫾ 2** 63† 90 ⫾ 3
LO to KP range
Hip (°) 30 ⫾ 8** 51 ⫾ 9 42 ⫾ 12** 56 ⫾ 11 65 ⫾ 14
Knee (°) 28 ⫾ 8* 41 ⫾ 11 37 ⫾ 14* 45 ⫾ 10 42 ⫾ 7
Trunk (°) 13 ⫾ 10 23 ⫾ 5 19 ⫾ 8 26 ⫾ 12 37 ⫾ 13
Thigh (°) 19 ⫾ 5 29 ⫾ 8 24 ⫾ 9 29 ⫾ 5 28 ⫾ 5
Shank (°) ⫺1 ⫾ 4** 7 ⫾ 3* 8 ⫾ 4** 4 ⫾ 3* 14 ⫾ 4
Minimum bar velocity (sticking point)
Hip (°) 111 ⫾ 22 123 ⫾ 17
Knee (°) 152 ⫾ 15* 164 ⫾ 8*
Trunk (°) 51 ⫾ 13 50 ⫾ 13
Thigh (°) 126 ⫾ 13** 107 ⫾ 6**
Shank (°) 77 ⫾ 5** 83 ⫾ 2**
Foot angle (°) 42 ⫾ 8** 14 ⫾ 6**
Significant differences between sumo and conventional deadlifts (** p ⬍ 0.01; * p ⬍ 0.05).
† Not statistically analyzed.

ferences observed in loads lifted, body mass and height, age, dorsiflexor moments exclusively, whereas the conventional
lift times, and Schwartz scores. Vertical bar and system group generated ankle plantar flexor moments exclusively.
displacements from LO to LC were significantly greater in Although at LO both groups generated knee extensor mo-
the conventional group. Furthermore, the total mechanical ments, at KP and LC the conventional group generated knee
work done on the barbell and system were significantly flexor moments, whereas the sumo group generated knee
greater in the conventional group. extensor moments. Although both groups generated hip
Select events and lifting phases are shown in Tables 4 and extensor moments, they were not significantly different from
5. Vertical bar velocity remained low throughout the lift. each other. Comparisons of ankle and knee moments and
Minimum bar velocity occurred at slightly less than 50% of moment arms between 2-D and 3-D analyses are shown in
the total lift time and slightly more than 50% of the total Table 7. For both the sumo and conventional deadlifts, 2-D
vertical bar distance. The conventional group reached the analyses were significantly different than 3-D analyses. The
first peak bar velocity significantly faster than the sumo largest differences observed occurred during the sumo deadlift.
group. Therefore, they spent significantly less time in the
acceleration phase than the sumo group. There were no
other significant differences observed. DISCUSSION
Joint resultant moments and moment arms are shown in Bilateral comparisons. In the current study, linear
Table 6. At LO, KP, and LC, ankle and knee moments and and angular displacements and velocities, as well as joint
moment arms were significantly different between sumo moments and moment arms, were averaged from the left and
and conventional groups. The sumo group generated ankle right sides of the body. There were no significant differ-
ences (P ⬍ 0.01) between bilateral measurements. Bilateral
TABLE 2. Comparisons of lower extremity joint and segment angles (mean ⫾ SD) measurements generally showed a 2–3° difference in joint
between two-dimensional (2-D) and three-dimensional (3-D) analyses. and segmental angles, a 2–3 cm difference in linear dis-
Sumo Conventional placements, and a 3– 4% difference in linear velocities.
2-D 3-D 2-D 3-D Bilateral differences in joint moment arms and moments
Analysis Analysis Analysis Analysis
(Fig. 2) were also minimal. The small differences observed
Liftoff
Hip (°) 88 ⫾ 14** 77 ⫾ 7** 75 ⫾ 14 72 ⫾ 12 between bilateral measurements demonstrate the symmetri-
Knee (°) 137 ⫾ 11** 126 ⫾ 8** 125 ⫾ 9 124 ⫾ 9 cal nature of the deadlift exercise.
Thigh (°) 127 ⫾ 9** 145 ⫾ 9** 132 ⫾ 8 134 ⫾ 7 Joint and segmental angles. Because there are no
Shank (°) 85 ⫾ 4** 80 ⫾ 4** 77 ⫾ 6 76 ⫾ 5
Knee passing 3-D analyses conducted on the deadlift, the objective of this
Hip (°) 121 ⫾ 12** 106 ⫾ 8** 122 ⫾ 15 114 ⫾ 11 study was to compare 3-D kinematics and kinetics between
Knee (°) 164 ⫾ 10** 153 ⫾ 8** 164 ⫾ 7 161 ⫾ 8
Thigh (°) 108 ⫾ 9** 126 ⫾ 9** 106 ⫾ 6 110 ⫾ 5 the sumo and conventional deadlifts. It was hypothesized
Shank (°) 93 ⫾ 2** 79 ⫾ 5** 91 ⫾ 4** 84 ⫾ 2** that joint and segment angles would show greater differ-
Significant differences between 2-D and 3-D analyses (** p ⬍ 0.01; * p ⬍ 0.05). ences between a 2-D versus 3-D analysis of the sumo
BIOMECHANICAL ANALYSIS OF THE DEADLIFT Medicine & Science in Sports & Exercise姞 1269
TABLE 3. Anthropometric, temporal, and work comparisons between sumo and conventional deadlifts.
Sumo (mean ⴞ SD) Conventional (mean ⴞ SD)
McGuigan and McGuigan and Brown and
Current Wilson (17) Current Wilson (17) Abani (2)
Age (yr) 47.4 ⫾ 7.3 46.4 ⫾ 6.1 17.9 ⫾ 1.1
Body height (cm) 172 ⫾ 8 170 ⫾ 8
Body mass (kg) 71.6 ⫾ 10.8 83.3 ⫾ 15.6 76.8 ⫾ 22.7 84.8 ⫾ 15.8 79.1 ⫾ 18.4
Barbell load (kg) 214.6 ⫾ 33.2 218.0 ⫾ 32.1 221.6 ⫾ 33.8 215.0 ⫾ 33.2 229.8 ⫾ 32.1
Schwartz Score (pts) 150.1 ⫾ 12.7 138.3 ⫾ 16.4 152.1 ⫾ 15.5 133.8 ⫾ 17.8

Stance width (cm) 70 ⫾ 11** 32 ⫾ 8**

Stance width (% shoulder width) 188 ⫾ 37** 80 ⫾ 16**
Hand width (cm) 47 ⫾ 4** 55 ⫾ 10**

Total lift time (s) 3.62 ⫾ 0.70 2.1 ⫾ 1.1 4.08 ⫾ 0.86 1.9 ⫾ 0.3
Time from LO to KP (s) 1.64 ⫾ 0.68 1.58 ⫾ 0.64
Time at sticking point (s) 1.68 ⫾ 0.54 1.77 ⫾ 0.50

Tot Vert Bar Dist (% Ht) 21.1 ⫾ 2.2** 26.8 ⫾ 4.2** 26.0 ⫾ 2.3** 32.9 ⫾ 2.3**
Tot Vert Sys Dist (% Ht) 18.6 ⫾ 2.6** 23.7 ⫾ 2.1**
Tot Vert Bar Dist (cm) 35.3 ⫾ 5.7** 44.4 ⫾ 5.7**
Tot Vert Sys Dist (cm) 32.1 ⫾ 5.2** 40.4 ⫾ 5.1**
Tot Mech Work on Bar (J) 739 ⫾ 139** 1003 ⫾ 268**
Tot Mech Work on Sys (J) 890 ⫾ 176** 1248 ⫾ 409**
Significant differences between sumo and conventional deadlifts (** p ⬍ 0.01; * p ⬍ 0.05).

deadlift compared with a 2-D versus 3-D analysis of the
conventional deadlift. Because during the sumo deadlift the
feet were approximately 70 cm apart and turned out approx-
imately 45° from the sagittal plane of the lifter, lower
extremity movements occurred in a plane midway between
the sagittal and frontal planes of the lifter in the direction the
feet were pointing, whereas trunk movements occurred in
the sagittal plane. Therefore, a 2-D analysis of the sumo
deadlift relative to the sagittal plane of the lifter produced
erroneous hip, knee, thigh, and shank measurements, espe-
cially the greater these movements deviated from the sag-
ittal plane. This is clearly demonstrated from Table 2, where
all joint and segment angles during the sumo deadlift were
significantly different between a 2-D and 3-D analysis.
Compared with a 3-D analysis, a 2-D analysis overestimated
hip, knee, and shank angles and underestimated thigh an-
gles. In contrast, during the conventional deadlift, measure-
ments from a 2-D analysis were generally not significantly
different than measurements from a 3-D analysis. During
the conventional deadlift the feet were turned out of the
sagittal plane only slightly. Therefore, body movements
primarily occurred in the sagittal plane. From these data, it
appears that a 2-D analysis is adequate in measuring joint
and segment angles during the conventional deadlift as
performed in the current study.
McGuigan and Wilson (17) conducted the only known
previous study that compared kinematic parameters between
the sumo and conventional deadlifts. Because their analyses
were 2-D, it was hypothesized that lower extremity mea-
surements between their study and the current study would
show greater differences than trunk measurements. In the
current study, hip, knee, thigh, and shank measurements
were generally significantly different between sumo and
conventional groups, whereas these same measurements
were generally not significantly different in McGuigan and
Wilson (17) (Table 1). Furthermore, compared with
McGuigan and Wilson (17), the current study found smaller
knee and hip angles and larger thigh angles. This is the same
pattern shown in Table 2 between a 2-D and 3-D analysis.
This implies that a 3-D analysis is needed to more accurately
calculate lower extremity measurements during the sumo
deadlift. In contrast, trunk measurements between the cur-
rent study and McGuigan and Wilson (17) produced the
same significant differences at LO, KP, and LC. This was
not surprising, because trunk movements between sumo and
conventional deadlifts occur primarily in the sagittal plane
of the lifter. Therefore, a 2-D analysis is adequate in quan-
tifying trunk angles.
Mechanical and physiological work. When normal-
ized by body height, the conventional group had 20 –25%
greater vertical bar and system center of mass (COM) dis-
placements from LO to LC compared with the sumo group
(Table 3). This is nearly identical to the vertical bar dis-
placements reported by McGuigan and Wilson (17). This is
not surprising because the sumo group had a stance width
2–3 times greater than the conventional group, which pro-

TABLE 4. Select events during the deadlift.

Sumo (mean ⴞ SD) Conventional (mean ⴞ SD)
Vertical Bar Time Vertical Bar
Velocity Time Occurred Position Velocity Occurred Position
Event (m䡠sⴚ1) (% total time) (% tot vert bar dist) (m䡠sⴚ1) (% total time) (% tot vert dist)
1st Peak bar velocity 0.205 ⫾ 0.063 26.1 ⫾ 10.0** 32.5 ⫾ 14.3 0.241 ⫾ 0.094 16.1 ⫾ 7.3** 26.0 ⫾ 14.0
Minimum bar velocity 0.076 ⫾ 0.560 46.5 ⫾ 14.9 56.4 ⫾ 24.9 0.091 ⫾ 0.084 43.3 ⫾ 12.2 61.6 ⫾ 19.1
2nd Peak bar velocity 0.150 ⫾ 0.087 65.0 ⫾ 12.3 76.1 ⫾ 14.1 0.163 ⫾ 0.095 61.6 ⫾ 16.2 80.4 ⫾ 11.3
Significant differences between sumo and conventional deadlifts (** p ⬍ 0.01; * p ⬍ 0.05).

1270 Official Journal of the American College of Sports Medicine http://www.msse.org

TABLE 5. Lifting phases during the deadlift.
Sumo
(mean ⴞ SD)
Acceleration phase (% total time) 26.1 ⫾ 10.0**
Sticking region (% total time) 20.4 ⫾ 10.3
Maximum strength region (% total time) 18.5 ⫾ 14.2
Deceleration phase (% total time) 35.0 ⫾ 12.3
Significant differences between sumo and conventional deadlifts (** p ⬍ 0.01; * p ⬍
0.05).
duced a lower COM. These lower COM displacements by
the sumo group resulted in 25–30% less mechanical work
performed compared with the conventional group.
Because the deadlift is considered a total body exercise,
with all the major muscles of the body being active, a high
energy expenditure results when performing this exercise.
Brown et al. (3) used standard open-circuit spirometry dur-
ing both active and recovery periods of the deadlift to
formulate the regression equation V̇O2 ⫽ 2.63 ⫹
0.80(Workbar), where V̇O2 is the total oxygen consumption
(L) and Workbar is the total mechanical work (kJ) done by
the lifter on the barbell during the deadlift ascent. These
authors formulated this equation to predict the total oxygen
cost while performing 1– 4 sets of 3–20 repetitions during
the deadlift. With a high correlation coefficient of 0.912,
83.2% of the variance between V̇O2 and Workbar was ex-
plained, which implies that the total estimated work during
the deadlift can be used as a gross predictor of the total
oxygen cost associated with mechanical work.
Powerlifters and other strength athletes in training typi-
cally perform the deadlift for 3– 4 sets of 3– 8 repetitions.
Consider an 80-kg athlete performing both the sumo and
convention deadlifts for 4 sets of 8 repetitions. Assume their
1 RM was 220 kg, and their total vertical bar displacement
was 0.444 m for the conventional deadlift and 0.353 m for
Conventional the sumo deadlift (data from Table 3). Because an 8-repe-
(mean ⴞ SD)
tition load is approximately 80% of the lifter’s 1 RM (16),
16.1 ⫾ 7.3**
27.2 ⫾ 11.3
the athlete will perform the 4 sets of 8 repetitions with a load
18.3 ⫾ 11.6 of approximately175 kg. Hence, the estimated total mechan-
38.4 ⫾ 16.2 ical work done during the 4 sets of 8 repetitions would be
24.4 kJ for the conventional deadlift and 19.4 kJ for the
sumo deadlift. Using the regression equation V̇O2 ⫽ 2.63 ⫹
0.80(Workbar) (3), the predicted total oxygen consumption
would be 22.1 L O2 during the conventional deadlift and
18.1 L O2 during the sumo deadlift. Because an oxygen
consumption of 1 L equates to an energy expenditure of
approximately 5 kcal, the 4 sets of 8 repetitions would
produce a predicted energy expenditure of approximately
111 kcal for the conventional deadlift and approximately 91
kcal for the sumo deadlift. These values are actually slightly
underestimated, because contributions from anaerobic gly-
colysis increase the total energy expenditure, and these
effects were only partially accounted for by the regression
equation. Because a typical 8 repetition deadlift set takes
approximately 30 s to complete, the total time needed to
perform 4 sets of 8 repetitions is approximately 2 min, plus
the rest time between sets. Assuming a 1.5-min rest period
between sets, which is similar to the rest period between sets
used in formulating the above regression equation, the total
time needed to perform 4 sets of 8 repetitions is approxi-
mately 6.5 min. This equates to an average rate of energy
expenditure of approximately 2.9 –3.5 L O2䡠min⫺1 (approx-
imately14 –17 kcal䡠min⫺1), which is relatively high com-
pared with walking, jogging, and other common activities
(1). The predicted energy expenditure during the deadlift is
even more impressive considering 70% of this time was

TABLE 6. Joint moments and moment arms (mean ⫾ SD) between sumo and conventional deadlifts.
Relative to Barbell Load COM Relative to System Load COM
Sumo Conventional Sumo Conventional
Moment arms at LO (cm)
Ankle ⫺17.2 ⫾ 6.8** 5.1 ⫾ 2.5** ⫺18.2 ⫾ 6.7** 4.2 ⫾ 2.7**
Knee ⫺17.8 ⫾ 5.0** ⫺3.9 ⫾ 3.6** ⫺19.3 ⫾ 5.0** ⫺5.1 ⫾ 3.6**
Hip 19.0 ⫾ 4.0 20.7 ⫾ 6.7 19.7 ⫾ 4.2 21.5 ⫾ 7.0
Moments at LO (N䡠m)
Ankle ⫺359 ⫾ 159** 109 ⫾ 50** ⫺503 ⫾ 211** 116 ⫾ 81**
Knee ⫺370 ⫾ 111** ⫺95 ⫾ 91** ⫺520 ⫾ 147** ⫺168 ⫾ 147**
Hip 403 ⫾ 121 461 ⫾ 164 512 ⫾ 154 599 ⫾ 230

Moment arms at KP (cm)

Ankle ⫺18.6 ⫾ 6.1** 1.4 ⫾ 3.5** ⫺19.7 ⫾ 6.1** 0.4 ⫾ 3.4**
Knee ⫺12.8 ⫾ 5.1** 2.2 ⫾ 2.7** ⫺14.1 ⫾ 5.2** 1.3 ⫾ 2.7**
Hip 17.4 ⫾ 4.0 16.2 ⫾ 4.5 18.2 ⫾ 4.2 16.9 ⫾ 4.7
Moments at KP (N䡠m)
Ankle ⫺389 ⫾ 145** 23 ⫾ 80** ⫺545 ⫾ 192** ⫺1 ⫾ 113**
Knee ⫺266 ⫾ 109** 46 ⫾ 66** ⫺381 ⫾ 146** 32 ⫾ 98**
Hip 366 ⫾ 102 351 ⫾ 78 469 ⫾ 133 457 ⫾ 98

Moment arms at LC (cm)

Ankle ⫺17.6 ⫾ 6.9** 2.9 ⫾ 3.4** ⫺19.5 ⫾ 6.8** 0.8 ⫾ 3.7**
Knee ⫺9.8 ⫾ 4.2** 2.4 ⫾ 3.5** ⫺11.8 ⫾ 4.2** 0.3 ⫾ 3.7**
Hip 8.7 ⫾ 2.9 6.0 ⫾ 2.0 7.0 ⫾ 3.9 4.0 ⫾ 2.0
Moments at LC (N䡠m)
Ankle ⫺369 ⫾ 167** 55 ⫾ 75** ⫺539 ⫾ 216** 5 ⫾ 121**
Knee ⫺204 ⫾ 93** 46 ⫾ 75** ⫺317 ⫾ 119** ⫺6 ⫾ 114**
Hip 190 ⫾ 92 132 ⫾ 45 190 ⫾ 135 110 ⫾ 53
Significant differences between sumo and conventional deadlifts (** p ⬍ 0.01; * p ⬍ 0.05).
LO, liftoff; KP, knee passing; LC, lift completion. Positive moment arms are anterior to joint; negative moment arms are posterior to joint; positive moments are hip extensor,
knee flexor, and ankle plantar flexor moments; negative moments are knee extensor and ankle dorsiflexor moments.

BIOMECHANICAL ANALYSIS OF THE DEADLIFT Medicine & Science in Sports & Exercise姞 1271
TABLE 7. Comparisons of ankle and knee joint moments and moment arms (mean ⫾ SD) relative to barbell load center of mass between two-dimensional (2-D) and three-
dimensional (3-D) analyses.
Sumo Conventional
2-D
Analysis 3-D Analysis 2-D Analysis 3-D Analysis
Moment arms at LO (cm)
Ankle 8.3 ⫾ 2.1** ⫺17.2 ⫾ 6.8** 9.2 ⫾ 2.4** 5.1 ⫾ 2.5**
Knee 4.8 ⫾ 1.9** ⫺17.8 ⫾ 5.0** ⫺0.1 ⫾ 2.8** ⫺3.9 ⫾ 3.6**
Moments at LO (N䡠m)
Ankle 175 ⫾ 49** ⫺359 ⫾ 159** 206 ⫾ 58** 109 ⫾ 50**
Knee 106 ⫾ 55** ⫺370 ⫾ 111** ⫺7 ⫾ 66** ⫺95 ⫾ 91**

Moment arms at KP (cm)

Ankle 6.3 ⫾ 1.1** ⫺18.6 ⫾ 6.1** 5.4 ⫾ 2.7** 1.4 ⫾ 3.5**
Knee 8.0 ⫾ 1.3** ⫺12.8 ⫾ 5.1** 6.1 ⫾ 1.4** 2.2 ⫾ 2.7**
Moments at KP (N䡠m)
Ankle 132 ⫾ 31** ⫺389 ⫾ 145** 116 ⫾ 55** 23 ⫾ 80**
Knee 170 ⫾ 44** ⫺266 ⫾ 109** 135 ⫾ 33** 46 ⫾ 66**

Moment arms at LC (cm)

Ankle 7.8 ⫾ 1.5** ⫺17.6 ⫾ 6.9** 7.0 ⫾ 2.7** 2.9 ⫾ 3.4**
Knee 7.7 ⫾ 2.4** ⫺9.8 ⫾ 4.2** 5.9 ⫾ 2.5** 2.4 ⫾ 3.5**
Moments at LC (N䡠m)
Ankle 162 ⫾ 34** ⫺369 ⫾ 167** 150 ⫾ 50** 55 ⫾ 75**
Knee 161 ⫾ 52** ⫺204 ⫾ 93** 127 ⫾ 44** 46 ⫾ 75**
Significant differences between 2-D and 3-D analyses (** p ⬍ 0.01; * p ⬍ 0.05).
LO, liftoff; KP, knee passing; LC, lift completion. Positive moment arms are anterior to joint; negative moment arms are posterior to joint; positive moments are knee flexor
and ankle plantar flexor moments; negative moments are knee extensor and ankle dorsiflexor moments.

spent resting. Most of the energy expended during the
deadlift occurred during the 2 min needed to perform the 4
sets of 8 repetitions, whereas a relatively smaller portion
was expended during the 4.5 min of rest. Therefore, total
body exercises like the deadlift not only enhance muscular
development and strength, but also generate a high energy
expenditure and have a high caloric cost. Furthermore, total
body multi-joint exercises are more functional in move-
ments of daily living and athletic endeavors compared with
single muscle, single joint exercises.
Selected events and lifting phases. Brown and
Abani (2) reported extremely low vertical bar accelerations
during the deadlift, with peak values 0.41 m䡠s⫺2 or less.
Similar low vertical bar accelerations were also observed in
the current study, as well as low vertical bar velocities
(Table 4). Three velocity events were noted during sumo
Figure 2—Representative graph for bilateral comparisons of joint
moments. Solid circle—right hip moment at barbell liftoff; open cir-
cle—left hip moment at barbell liftoff.
1272 Official Journal of the American College of Sports Medicine
and conventional deadlifts: a) 1st peak bar velocity; b)
minimum bar velocity; and c) 2nd peak bar velocity. The
typical vertical bar velocity pattern (Fig. 3) during the dead-
lift was that the bar initially accelerated from LO to 1st peak
bar velocity, slowed down to minimum bar velocity, again
accelerated to 2nd peak bar velocity, and finally slowed
down as LC approached. The COM of the bar actually
began moving vertically before the barbell disks leaving the
ground at liftoff (Fig. 3). The bar bent slightly as the lifter
initially exerted an upward force on the bar with his hands,
which occurred because the bar is not completely rigid.
Nevertheless, the amount of bend in the bar was minimal.
The vertical bar velocity curve shown in Figure 3 is the
same general pattern observed during maximum or near
maximum loads during the squat (18), but different than the
vertical bar velocity curve presented by McGuigan and
Wilson (17). This discrepancy in the deadlift may be due the
fact that the lifters in McGuigan and Wilson (17) completed
the deadlift in approximately half the time it took the lifters
in the current study (Table 3). The approximately 4-s total
lift time in the current study is in agreement with another
study involving the 1 RM deadlift during a national pow-
erlifting championship (5). This implies that the subjects in
McGuigan and Wilson (17) may have underestimated their
1 RM. Although powerlifters in competition typically
choose a weight that they believe represents their 1 RM, this
weight is sometimes less than their 1 RM. This may be
because they underestimate their 1 RM, or it may be a
strategic move in order to defeat their opponent. In either
case, the amount of weight they lift may actually be closer
to 95% of their 1 RM, which will affect both lifting kine-
matics and kinetics. Because the lifters in the current study
and in Cholewicki et al. (5) were national level powerlifters,
they may have been more experienced than the lifters in
McGuigan and Wilson (17), who competed in regional
http://www.msse.org

Figure 3—Representative graph for vertical bar velocity. AP, accel-
eration phase; SR, sticking region; MSR, maximum strength region;
DP, deceleration phase.
powerlifting competitions. A more experienced lifter can
better predict their 1 RM in competition compared to a less
experienced lifter.
Of the 24 lifters in the current study, 18 lifters achieved
maximum vertical bar velocity at their 1st peak vertical bar
velocity, whereas the remaining six lifters achieved maxi-
mum vertical bar velocity at their 2nd peak vertical bar
velocity. From Table 4, the mean 1st peak bar velocity was
approximately 40 –50% greater than the mean 2nd peak bar
velocity. These data are different compared to a 1 RM squat
(18), in which lifters typically reached their maximum ver-
tical bar velocity at their 2nd peak vertical bar velocity.
During the 1 RM squat, the mean 2nd peak bar velocity was
approximately 40 –50% greater than the mean 1st peak bar
velocity (18), which is opposite the findings from the cur-
rent study. The maximum vertical bar velocity during the
powerlifting squat had a mean value of approximately 0.5
m䡠s⫺1 (18), which is approximately twice the mean value of
0.238 ⫾ 0.083 m䡠s⫺1 observed in the current study. Maxi-
mum vertical bar velocities in the current study ranged from
0.108 – 0.456 m䡠s⫺1, which were less than the 0.580 m䡠s⫺1
reported by McGuigan and Wilson (17) for one subject
performing the deadlift. This is not surprising since their
subjects completed the deadlift in half the time as the
subjects in the current study.
As originally defined by Lander et al. (13), the lift was
divided into four phases (Table 5, Fig. 3): a) acceleration
phase 3 LO to 1st peak bar velocity; b) sticking region 3
1st peak bar velocity to minimum bar velocity; c) maximum
strength region 3 minimum bar velocity to 2nd peak bar
velocity; and d) deceleration phase 3 2nd peak bar velocity
to LC. The acceleration phase and sticking region for the
conventional group consisted of approximately 16% and
27%, respectively, of the total lift time. This was different
from the sumo group, who spent approximately10% more
time in the acceleration phase (Table 5). Interestingly, ap-
proximately 18% of the total lift time was spent in the
maximum strength region during the deadlift, whereas ap-
proximately 35–38% of the total lift time was spent in the
deceleration phase.
BIOMECHANICAL ANALYSIS OF THE DEADLIFT
The end of the sticking region has previously been re-
ported as the “sticking point” (18). Because bar velocity is
minimal at this instant, it appears to be the most difficult part
of the lift. This is often where powerlifters fail in their
attempt for a successful lift. In the current study, the sticking
point occurred just after KP (Table 3), occurring at approx-
imately 43% of the total lift time for the conventional group
and approximately 47% of the total lift time for the sumo
group (Table 4). These are similar to the relative times
reported by McGuigan and Wilson (17). The sticking point
phenomena may in part be due to mechanical principles of
skeletal muscle, such as to the length-tension relationship.
The greater a muscle shortens, the less tension it is capable
of generating. Decreasing muscle moment arm lengths from
KP to LC may also be responsible for the occurrence of the
sticking point. The sticking point occurred at a slightly
different body position compared with a 1 RM squat (18).
The trunk was slightly more upright in the deadlift (approx-
imately 50°) compared with the squat (approximately 39°),
and the knees and hips were extended more in the deadlift
(approximately 152–164° and 111–123°, respectively) com-
pared with the squat (approximately 105 ⫾ 6° and 70 ⫾ 11°,
respectively). These body position differences between the
squat and deadlift occur in part because the barbell is behind
the lifter during the squat and in front of the lifter during the
deadlift.
Joint moments and moment arms. Cholewicki et
al. (5) conducted the only known kinetic comparison be-
tween sumo and conventional deadlifts. Their subjects lifted
approximately the same amount of weight as the subjects in
the current study. Because these investigators used only one
camera in recording a sagittal view of the lifters, they
performed a 2-D analysis and calculated moments and mo-
ment arms at LO about the lumbar spine, hip, and knee.
Although a 2-D analysis is appropriate in calculating hip and
spinal moments and moment arms during both the sumo and
conventional deadlifts, it is only appropriate in calculating
ankle and knee moments and moment arms if the feet are
pointing straight ahead. Because the feet are turned out to a
greater extend in the sumo deadlift (42°) compared with the
conventional deadlift (14°), a 2-D analysis for the sumo
deadlift will produce greater errors in ankle and knee mo-
ments and moment arms compared with the conventional
deadlift. This is clearly demonstrated in Table 7. Although
2-D analyses were significantly different than 3-D analyses
for both the sumo and conventional deadlifts, the largest
differences occurred during the sumo deadlift. From Table
7, ankle and knee moment arms for the conventional deadlift
were generally only a few centimeters different between a
2-D and 3-D analysis. In contrast, ankle and knee moment
arms for the sumo deadlift were generally 20 –25 cm dif-
ferent between a 2-D and 3-D analysis.
Similar to data from Cholewicki et al. (5), there were no
significant differences in hip moments and moment arms
between sumo and conventional deadlifts. However, unlike
the knee moments in the current study (Table 6), knee
moments from Cholewicki et al. (5) were not significantly
different between sumo (18 N䡠m) and conventional (18
Medicine & Science in Sports & Exercise姞 1273
N䡠m) deadlifts. The knee moments calculated from a 2-D
analysis (Table 7) are relatively similar in magnitude to the
2-D knee moments from Cholewicki et al. (5) but quite
different than the knee moments calculated by a 3-D anal-
ysis (Table 7). Ankle moments have previously been re-
ported only for the conventional deadlift (2), with these
authors finding similar ankle moment magnitudes as the
current study.
Joint moments and moment arms during the deadlift and
squat have been reported with respect to the barbell COM
only (2,5,19). Because joint moments are also generated by
body segments, moments and moment arms relative to the
system load (i.e., barbell plus those body segments that
contribute to generating joint moments) were also calculated
(Table 6). Although moment arms were generally only 1–2
cm different between the barbell COM and the system
COM, system-generated joint moments were much greater
than barbell generated joint moments due to the greater load
of the system. During the sumo deadlift, ankle and knee
moments were generally 40 –50% greater with respect to the
system weight compared with the barbell weight. Similarly,
hip moments at LO and KP were generally 25–30% greater
for the system weight compared with the barbell weight.
Therefore, joint moment contributions from body segments
should not be discounted when calculating the actual joint
moments that occur during lifting.
One of the most interesting findings in the current study
is that ankle plantar flexor moments and knee flexor mo-
ments were primarily generated during the conventional
deadlift, whereas ankle dorsiflexor moments and knee ex-
tensor moments were generated during the sumo deadlift. In
both deadlift styles, hip extensor moments were generated
exclusively. This suggest that during the conventional dead-
lift the ankle plantar flexors, knee flexors, and hip extensors
are primarily responsible for causing or controlling move-
ments at the ankles, knees, and hips, respectively. Similarly,
during the sumo deadlift the ankle dorsiflexors, knee exten-
sors, and hip extensors are primarily responsible for causing
or controlling ankle, knee, and hip movements, respectively.
It can be hypothesized from these data that the primary
lower extremity muscles involved during the conventional
deadlifts are the hamstrings, gluteus maximus, gastrocne-
mius, and soleus, whereas the primary lower extremity
muscles involved during the sumo deadlifts are the gluteus
maximus, hamstrings, quadriceps, and tibialis anterior.
Electromyography (EMG) should be employed during these
two deadlift styles to test these hypotheses.
It should be noted that ankle and knee moments during
the sumo deadlift are relatively large in magnitude com-
REFERENCES
1. AMERICAN COLLEGE OF SPORTS MEDICINE. Guidelines for Exercise
Testing and Prescription, 4th Ed., R. R. Pate (Ed.).. Philadelphia:
Lea & Febiger, pp. 285–300, 1991.
2. BROWN, E. W., and K. ABANI. Kinematics and kinetics of the dead lift
in adolescent power lifters. Med. Sci. Sports Exerc. 17:554–566, 1985.
3. BROWN, S. P., J. M. CLEMONS, Q. HE, and S. LIU. Prediction of the
oxygen cost of the deadlift exercise. J. Sports Sci. 12:371–375, 1994.
1274 Official Journal of the American College of Sports Medicine
pared with ankle and knee moments generated during the
conventional deadlift. For example, at KP the system load
COM is slightly anterior to the knee axis during the
conventional deadlift (Table 6), thus generating a very
small knee extensor moment. Consequently, the knee
extensor moment due to the external barbell load attempts
to extend the knees. To counteract this knee extensor
moment, increased activity from the knee flexors gener-
ate a knee flexor moment needed in order to control the
rate and extent of knee extension. This is one reason why
the hamstrings may be more active and the quadriceps
less active during the conventional deadlift compared
with the sumo deadlift. Because the knee flexors are also
hip extensors, they are attempting to extend the hip and
flex the knee simultaneously. In contrast, because the
system load COM is posterior to the knee axis during the
sumo deadlift, a large knee extensor moment is needed to
counteract the large knee flexor moment generated by the
system load. This implies that the quadriceps may be
more active during the sumo deadlift compared with the
conventional deadlift.
If the external barbell load causes the knees to extend
prematurely or excessively, the lifter will have to complete
what is referred to as a “stiff-leg” deadlift, which occurs
when the knees are at or near full extension. This decreases
quadriceps activity, increases hamstring activity (28), and
increases erector spinae activity due to the spine being in a
more bent and rounded position. This also places the lumbar
spine in a more vulnerable position, with a higher risk of
injury (5). Although spinal forces and moments were not
quantified, the more upright trunk position observed during
the sumo deadlift may decrease spinal loads. Cholewicki et
al. (5) have shown that compared with the conventional
deadlift, the sumo deadlift generated a 10% reduction in the
L4/L5 moment and an 8% decrease in the L4/L5 shear force.
This suggests that the lower back may be at lower risk of
injury during the sumo deadlift. However, the conventional
deadlift may be more effective in strengthening back and
hamstring musculature. An EMG analysis is needed to test
these hypotheses.
The authors extend a special thanks to Andy Demonia, for all his
help and support in collecting the data, and a special thanks to Tom
and Ellen Trevorah (meet directors), for all their support for this
project.
Address for correspondence: Rafael Escamilla, Ph.D., C.S.C.S.,
Duke University Medical Center, P.O. Box 3435, Durham, NC 27710;
E-mail: rescamil@duke.edu.
4. CHOLEWICKI, J., and S. M. MCGILL. Lumbar posterior liga-
ment involvement during extremely heavy lifts estimated
from fluoroscopic measurements. J. Biomech. 25:17–28,
1992.
5. CHOLEWICKI, J., S. M. MCGILL, and R. W. NORMAN. Lumbar spine
loads during the lifting of extremely heavy weights. Med. Sci.
Sports Exerc. 23:1179 –1186, 1991.
http://www.msse.org
 6. DEMPSTER, W. T. Space requirements of the seated operator
(WADC Technical Report). Wright-Patterson Air Force Base, OH,
pp. 55–159, 1955.
7. ELLIOTT, B. C., G. J. WILSON, and G. K. KERR. A biomechanical
analysis of the sticking region in the bench press. Med. Sci. Sports
Exerc. 21:450 – 462, 1989.
8. GRANHED, H., R. JONSON, and T. HANSSON. The loads on the lumbar
spine during extreme weight lifting. Spine. 12:146 –149, 1987.
9. HARMAN, E. A., P. N. FRYKMAN, E. R. CLAGETT, and W. J. KRAE-
MER. Intra-abdominal and intra-thoracic pressures during lifting
and jumping. Med. Sci. Sports Exerc. 20:195–201, 1988.
10. HARMAN, E. A., R. M. ROSENSTEIN, P. N. FRYKMAN, and G. A.
NIGRO. Effects of a belt on intra-abdominal pressure during weight
lifting. Med. Sci. Sports Exerc. 21:186 –190, 1989.
11. HENNING, C. E., M. A. LYNCH, and K. R. GLICK, JR. An in vivo
strain gage study of elongation of the anterior cruciate ligament.
Am. J. Sports Med. 13:22–26, 1985.
12. LANDER, J. E., B. T. BATES, and P. DEVITA. Biomechanics of the
squat exercise using a modified center of mass bar. Med. Sci.
Sports Exerc. 18:469 – 478, 1986.
13. LANDER, J. E., B. T. BATES, J. A. SAWHILL, and J. HAMILL. A
comparison between free-weight and isokinetic bench pressing.
Med. Sci. Sports Exerc. 17:344 –353, 1985.
14. LANDER, J. E., R. L. SIMONTON, and J. K. GIACOBBE. The effective-
ness of weight-belts during the squat exercise. Med. Sci. Sports
Exerc. 22:117–126, 1990.
15. LUTZ, G. E., R. A. PALMITIER, K. N. AN, and E. Y. CHAO. Com-
parison of tibiofemoral joint forces during open-kinetic-chain and
closed-kinetic-chain exercises. J. Bone Joint Surg. Am. 75:732–
739, 1993.
16. MAYHEW, J. L., J. R. WARE, and J. L. PRINSTER. Using lift repeti-
tions to predict muscular strength in adolescent males. Natl.
Strength Condit. J. 15:35–38, 1993.
17. MCGUIGAN, M. R. M., and B. D. WILSON. Biomechanical analysis
of the deadlift. J. Strength Condit. Res. 10:250 –255, 1996.
BIOMECHANICAL ANALYSIS OF THE DEADLIFT
18. MCLAUGHLIN, T. M., C. J. DILLMAN, and T. J. LARDNER. A kine-
matic model of performance in the parallel squat by champion
powerlifters. Med. Sci. Sports. 9:128 –133, 1977.
19. MCLAUGHLIN, T. M., T. J. LARDNER, and C. J. DILLMAN. Kinetics of
the parallel squat. Res. Q. 49:175–189, 1978.
20. MORE, R. C., B. T. KARRAS, R. NEIMAN, D. FRITSCHY, S. L. WOO,
and D. M. DANIEL. Hamstrings–an anterior cruciate ligament pro-
tagonist: an in vitro study. Am. J. Sports Med. 21:231–237, 1993.
21. NISELL, R., and J. EKHOLM. Joint load during the parallel squat in
powerlifting and force analysis of in vivo bilateral quadriceps
tendon rupture. Scand J. Sports Sci. 8:63–70, 1986.
22. OHKOSHI, Y., K. YASUDA, K. KANEDA, T. WADA, and M. YA-
MANAKA. Biomechanical analysis of rehabilitation in the standing
position. Am. J. Sports Med. 19:605– 611, 1991.
23. PALMITIER, R. A., K. N. AN, S. G. SCOTT, and E. Y. CHAO. Kinetic
chain exercise in knee rehabilitation. Sports Med. 11:402– 413, 1991.
24. SHAPIRO, R. Direct linear transformation method for three-dimen-
sional cinematography. Res. Q. 49:197–205, 1978.
25. SHELBOURN, K. D., and P. NITZ. Accelerated rehabilitation after
anterior cruciate ligament reconstruction. Am. J. Sports Med. 18:
292–299, 1990.
26. STUART, M. J., D. A. MEGLAN, G. E. LUTZ, E. S. GROWNEY, and
K. N. AN. Comparison of intersegmental tibiofemoral joint forces
and muscle activity during various closed kinetic chain exercises.
Am. J. Sports Med. 24:792–799, 1996.
27. WOOD, G. A., and R. N. MARSHALL. The accuracy of DLT extrap-
olation in three-dimensional film analysis. J. Biomech. 19:781–
785, 1986.
28. WRIGHT, G. A., T. H. DELONG, and G. GEHLSEN. Electromyographic
activity of the hamstrings during performance of the leg curl,
stiff-leg deadlift, and back squat movements. J. Strength Condit.
Res. 13:168 –174, 1999.
29. YACK, H. J., C. E. COLLINS, and T. J. WHIELDON. Comparison of
closed and open kinetic chain exercise in the anterior cruciate
ligament-deficient knee [see comments]. Am. J. Sports Med. 21:
49 –54, 1993.
Medicine & Science in Sports & Exercise姞 1275