Authors
Erica M. Buckeridge 1,2 ,3
,
Anthony M.J. Bull 2
,
Alison H. McGregor 1
1
Department of Surgery & Cancer, Imperial College London, UK
2
Department of Bioengineering, Imperial College London, UK
3
Human Performance Laboratory, University of Calgary, Canada
Corresponding Author:
Erica Buckeridge
1,2 ,3
Human Performance Laboratory, Room B225
Faculty of Kinesiology, University of Calgary
Calgary, AB, Canada, T2N 1N4
Phone: +01(403) 220-3449 ebuckeridge@kin.ucalgary.ca
Running Title: Rowing technique and performance
Abstract: 227 words
Main text: 4644 words
1

Abstract
Rowing is a sport in which the parameters of injury, performance and technique are all interrelated and in dynamic equilibrium. Therefore, whilst rowing requires extreme physical strength and endurance, a high level of skill and technique is essential to enable an effective transfer of power through the rowing sequence. This study aimed to determine discrete aspects of rowing technique which strongly influence foot force production and asymmetries at the foot stretchers, as these are biomechanical parameters often associated with performance and injury risk.
Twenty elite female rowers participated in this study. A motion analysis system was used with a rowing ergometer which was instrumented to measure force at the handle and foot stretchers.
During an incremental rowing test, three-dimensional kinematic recordings of the ankle, knee, hip and lumbar pelvic joints were made, in addition to measures of force at the handle and foot stretchers. Multiple regression analyses identified hip kinematics as a key predictor of foot force output (R
2
=.48), whereas knee and lumbar-pelvic kinematics were the main determinants in optimising the horizontal foot force component (R
2
=.41). Bilateral asymmetries of the foot stretchers were also seen to significantly influence lumbar-pelvic kinematics (R
2
=.43) and pelvic twisting (R
2=
.32) during the rowing stroke. These results provide biomechanical evidence towards aspects of technique that can be modified optimize force output and performance, which can be of direct benefit to coaches and athletes.
Key words: kinematics, foot force, asymmetry, regression
2

Introduction
In the Olympic sport of rowing, high mean velocity of the boat can be achieved through the production of large, symmetrical foot forces which are efficiently delivered through the human
kinetic chain to the handle/oars (Hofmijster et al., 2008). Suboptimal technique through
inaccurate sequencing of body segment motion negatively affects the forces applied to the footstretcher, and impacts the efficiency of transfer to the handle/oars, thus reducing mean boat
velocity (Baudouin and Hawkins, 2002). For example, ‘shooting the slide’ is when the legs
extend to move the seat posteriorly, but the motion is disconnected from the trunk and handle,
thus resulting in a miss-timed trunk rotation that wastes energy from the leg drive (McNeely and
Royle, 2002). Other common technical faults include taking the catch with the shoulders and
upper body instead of engaging the leg drive, and over extending at the finish thus weakening
lumbar-pelvic posture (Bull and McGregor, 2000). Consequently, the movement patterns which
occur during and following the leg drive are important factors in optimising power per stroke.
Thus from a coaching perspective, specific kinematic variables which influence the production of large foot forces are worth investigating.
More recently, there has been an emphasis on examining the ability to apply mechanically effective forces i.e. quantifying the horizontal component of force relative to the resultant force
(Caplan and Gardner, 2010, Morin et al., 2011). Although the studies described below did not
directly measure foot force, it has previously been shown that raising foot-stretcher heights compared to a standard height results in a more posterior trunk motion and improved power output. With regards to rowing, it was previously suggested that increased height of the footstretchers placed the rower in a more mechanically effective position to generate horizontal foot
force, which subsequently improved power output (Caplan and Gardner, 2005, Caplan and
3

Gardner, 2010). Furthermore, research into sprinting has found a key performance predictor to
be horizontal impulse during the stance phase (Hunter et al., 2005). In agreement, Morin et al.
(2011) suggested that it was not the total force applied during the stance phase of a sprint, but
rather the proportion of horizontal force with respect to the resultant force, which determined sprint performance. This was termed the foot force ratio (RF). Thus, for the same overall force magnitude, different strategies of force application may be employed that result in larger horizontal force output and thus, net forward accelerations. Much like in sprinting, rowing requires acceleration in the horizontal plane to be maximised, and as such, it is important to examine discrete aspects of rowing technique which optimise RF.
In addition to the magnitude and orientation of foot stretcher forces, the importance of a symmetrical leg drive cannot be overlooked. Rowing is often studied unilaterally from a
biomechanical perspective (Halliday et al., 2004, Hase et al., 2004), although recent studies have
shown evidence of lower limb asymmetries. For example, asymmetries have been identified at the hips and knees, with those at the hips proving to be a moderate predictor of lumbar-pelvic
flexion during the drive phase (Buckeridge et al., 2012). Asymmetrical loading of the footplates has received limited attention to date (Baca et al., 2006). Furthermore, its influence on motion
further up the kinematic chain, such as pelvic twist and lumbar-pelvic motion, has not been examined. Given that rowers are constrained at the hands, seat and feet and are performing a closed chain activity, any asymmetrical loading of the foot-stretchers must be balanced through compensatory movement patterns or stabilizing co-contractions in order to keep the motion in the sagittal plane. For example asymmetrical lifting has been shown to cause asymmetrical torques at the lower back, which result in increased spinal loading through stabilizing co-
contractions of the lumbar spine (Kingma et al., 1998). Movements of the pelvis which
4

counteract asymmetries will, in theory, impact the action of the lumbar spine and also create out of plane motion artefacts on the seat that are deleterious to performance. With high importance of technique in the optimisation of the rowing stroke, the possible influence of asymmetrical foot stretcher loading on technique could be a limiting factor to performance. As such, it is important to investigate the possible impact of asymmetrical loading of the footplates on subsequent kinematics of the rower.
Clearly there is an important interaction in cyclic activities where technique affects performance output and vice versa. Therefore, this study will aim to investigate multiple interactions between performance and technique by; (a) investigating discrete aspects of technique that make significant contributions to large resultant foot force, (b) investigating discrete aspects of technique that elicit large RF at key points during the stroke, (c) examine whether asymmetries at the feet influence twisting of the pelvis and flexion of the lumbar-pelvic joint. It is hypothesised that small changes in rowing technique through altered lower limb and lumbar-pelvic kinematics have the capacity to influence force production and asymmetries at the foot stretchers.
METHODS:
Participants
The study received ethical approval from the Imperial College Research Ethics Committee, and informed consent was obtained from all subjects. Twenty heavyweight female rowers from the
GB rowing squad participated in this study (weight 76.6 ± 5.1 kg, age 26.8 ± 3.0). As the study has a focus on symmetry, subjects with known geometry asymmetry, that is a leg difference of more than 1 cm, and rowers with a current episode of low back pain or any other serious illness or injury were excluded.
5

Hardware
All rowing trials were performed on a modified Concept2 model D ergometer (Concept2 Inc.,
Morrisville, Vermont, USA). The ergometer was instrumented at the handle with a uniaxial load cell (ELHS model, Entran, Lexington, Kentucky, USA) to measure pulling force on the handle
(2.5 kN range, 0.5% combined non-linearity and hysteresis) (Loh et al., 2004, McGregor et al.,
2008). Foot-stretchers were replaced with strain gauge instrumented footplates (Vishay, Micro
Measurements, UK) which bilaterally measure vertical, horizontal and their resultant foot forces
(Buckeridge et al., 2013).
Right side lower limb and lumbar-pelvic kinematics of rowers were recorded using the Flock of
Birds (FOB) motion capture system (Ascension Technology, Burlington, VT). The system comprised of an electromagnetic extended range transmitter, positioned 1.0 m laterally from the ergometer and 1.5 m off the ground, and four receivers (S1 – S4) which could quantify six degrees of freedom kinematics within the electromagnetic field. Previous work has validated
this system’s suitability for measuring spinal and lower limb motion (Bull et al., 1998, Bull and
McGregor, 2000, Bull et al., 2004).
The data acquisition PC communicated with the FOB system via a serial port. Signals from the instrumented ergometer were connected to the same PC through a signal conditioning unit (SC-
2345, National Instruments, Austin, Texas, USA), and the PC interface for this unit was through a PCI card (DAQ PCI-5036E, National Instruments, Austin, Texas, USA). A custom program was developed in Labview (Version 2009, National Instruments, Austin, Texas, USA) for software synchronisation of all FOB and ergometer output signals, data acquisition, and real-time biofeedback.
6

Subject digitization
A modified version of the sensor set-up and digitization procedure outlined in Buckeridge et al.
(2012) was utilised. Figure 1 depicts the placement of each sensor and the segment it represents.
Adhesive pads (PALstickies
TM
, PAL Technologies Ltd, Glasgow, Scotland) were used to secure
S1 and S2 to the skin, whilst S4 was attached to a foam cuff and strapped to the subjects’ right shank. S3 was employed as a stylus for digitization of bony landmarks, prior to being fixed to the thigh with Velcro straps. Whilst the rower was seated on the ergometer, the tip of the stylus was placed on landmarks of interest and rotated about that point in order to create a cloud of 3D position data. A sphere fitting procedure was then used to work out the 3D vector offset of that point relative to the sensors already attached to the body segments, so that the trajectories of the landmarks could be tracked at all points during the rowing stroke. Table 1 shows the bony landmarks that are digitised, and indicates the sensors that they are stored as offsets from. The hip joint centre was determined through a functional test. This involved strapping the S3 digitising stylus to the subject’s thigh, and in a standing position, the subject rotated their thigh to capture their full range of motion about the hip, enabling a sphere fitting procedure to find the centre of rotation.
Protocol
After a 10-minute warm up each athlete performed an incremental step test on the modified
Concept2 ergometer. The Step Test involved rowing for three minutes at each of the following stroke rates:

18 strokes per minute at an effort equivalent to 55% of their 2000 m personal best.

24 strokes per minute (no split given).
7


28 strokes per minute (no split given).

Free rate i.e. an arbitrary stroke rate that corresponds to rowing a 500 m split at race pace
(McGregor et al., 2008).
Sufficient rest was permitted between each step. All rowers had previously performed this step test on multiple previous occasions, thus were familiar with the protocol.
Three dimensional kinematics model
A kinematics model identified five joint centres: lumbar-pelvic joint (L5/S1) defined as a local offset from S2, hip joint centre (HJC) defined using the functional digitisation method, knee joint centre (KJC) defined as the mid-point between the MEPI and LEPI, ankle joint centre (AJC) defined as the mid-point between the MMAL and LMAL, and foot joint centre (FJC) defined as an offset from the dorsal aspect of the fifth metatarsal head, coincident with the global X-axis.
Local co-ordinate frames of five segments are defined in the model, based on 3D vectors connecting digitised landmarks, as described by Murphy (2009). These included the lumbar spine segment, pelvis segment, thigh segment, shank segment and foot segment. Joint coordinate frames for L5/S1, HJC, KJC and AJC were quantified using the method described by
Grood and Suntay (1983), and this subsequently enabled three-dimensional angles
(flexion/extension, abduction/adduction and internal/external rotation) of each joint to be quantified. For all joints, 0° is full joint extension, whilst knee flexion is positive and L5/S1, hip and ankle flexion are negative. Abduction and external rotation of joints were also defined as positive.
Data processing
8

Force outputs from the foot stretchers were recorded in Newtons (N) and subsequently normalised to the rowers’ body mass in kilograms. Vertical and horizontal foot forces were recorded directly from the right and left footplates, enabling the resultant force to be quantified per footplate.
Bilateral foot force asymmetries were quantified using the absolute version of the symmetry index
(Robinson et al., 1987), as shown in equation 1:
𝐴𝑆𝐼 (%) =
2|𝑋 𝑟𝑖𝑔ℎ𝑡
−𝑋 𝑙𝑒𝑓𝑡
|
∙ 100
(𝑋 𝑟𝑖𝑔ℎ𝑡
+𝑋 𝑙𝑒𝑓𝑡
)
Equation 1
X right is the value of the right limb and X left is the value of the left limb. An ASI value of zero indicates perfect symmetry, and increasingly positive values indicate increasing magnitudes of bilateral asymmetry.
Bilateral foot forces were summed to give total resultant, vertical and horizontal foot force values. Average resultant foot force was derived for the drive phase (catch – finish).
Furthermore, based on the total vertical and horizontal components of force, the foot force ratio
(RF) was derived. This represents the effectiveness of force application in the horizontal plane
(Morin et al., 2011).
𝑅𝐹 =
𝐹 ℎ𝑜𝑟𝑖𝑧
𝐹 𝑟𝑒𝑠
Equation 2
Where RF is foot force ratio, F horiz is horizontal foot force and F res
is resultant foot force.
All kinetic and kinematic variables were extracted at the catch, MHF (maximum handle force) and finish positions for ten consecutive time normalised (0-100%) strokes in the middle of each 3 minute piece. The catch at 0% was defined as the onset of tensile force at the handle where the force
9

first exceeds 75 N, with previous studies finding this method to be a repeatable measure of the
catch (Holt et al., 2003, McGregor et al., 2005). The finish was the point at which tensile force
production was less than 50 N and MHF was the point at which peak handle force occurred.
Foot force ratio was extracted at MHF only, as it is important to know what the foot force orientation is at a point during the high loading phase of the rowing stroke. Additionally, joint ROM in the sagittal plane was calculated for each joint, with ROM representing the motion that occurred between maximum and minimum values that occurred per stroke. A key variable pertaining to posture was the change in sagittal plane angle of L5/S1 between catch and MHF (ΔL5/S1). Pelvic twist was calculated based on the positions of the left and right ASIS landmarks with respect to the vector
between the MET5 landmarks (Buckeridge et al., 2012). A positive angle was defined as
clockwise pelvic twist, and a negative angle was defined as anti-clockwise pelvic twist.
Statistical analysis
To examine the relationships between kinematic variables and foot force, Stepwise multiple linear regressions were performed. Where average resultant foot force and RF at MHF were dependent variables in two separate multiple regression models, explanatory variables for those models were the Kinematic Explanatory Variables listed in Table 2. Where ΔL5/S1 and pelvic twist were dependent variables in a further two regression models, explanatory variables for those models were the ASI Explanatory Variables in Table 2. The criterion for the independent variable’s entry into the multiple regression model was p <0.05. The quality of the regression model was evaluated through post-test assessments of R
2
and adjusted R
2
values. All statistical analyses were performed using SPSS (version 19, IBM Corporation, New York, USA).
10

Diagnostics were performed to test two important suppositions: that the model fits the observed data well, and that the model is generalizable to a larger sample. To satisfy the first supposition, cases where standardized residuals had z-scores greater than 3.0 (i.e. three standard deviations
from the mean) were treated as outliers and removed from the model (Field, 2009). Influential
cases were additionally examined through values of Cook’s distance, mean leverage and
Mahalanobis distance to ensure individual cases had only a small influence on the regression line.
To enable generalizable conclusions to be drawn from this model, assumptions of homoscedasticity, normally distributed errors, independent errors and multicollinearity were checked to ensure they were not violated. Independent errors were checked with the Durbin
Watson statistic, where 2.0 indicates zero correlation, and multicollinearity was checked with a variance inflation factor, where values greater than 10.0 is considered problematic and thus excluded.
Results
The relationship between resultant foot force and lower limb kinematics
The first aim was to examine the relationship between lower limb kinematics and average resultant foot force. The mean and standard deviations of the dependent variable, in addition to the independent variables which were extracted by the Stepwise regression model are shown in Table 3.
A positive regression coefficient tells you that an increase in the independent variable’s value results in increasing values of the dependent variable.
Three predictors were significantly correlated with average resultant force; MHF hip flexion, hip
ROM and catch ankle flexion . With these variables, the model achieved an R
2 and adjusted R
2 value of 0.48 and 0.46 respectively. Catch hip flexion was originally a fourth variable which
11

satisfied the model criteria, but was removed from the Stepwise regression as its correlation with
MHF hip flexion was very high ( R =0.81) and thus violated the assumption of multicollinearity.
The relationship between foot force ratio and lower limb kinematics
The second aim was to determine the discrete lower limb kinematic variables which influence RF.
The mean and standard deviations of the dependent variable, in addition to the independent variables which were extracted by the Stepwise regression model are shown in Table 4. Only two dependent variables;
ΔL5/S1 and MHF knee flexion were extracted, with the model achieving an R 2
and adjusted R
2
value of 0.41 and 0.39 respectively.
The relationship between foot force asymmetries and lumbar-pelvic kinematics
The third aim of this study was to investigate whether asymmetries at the footplate influence, and could thus be used as a surrogate measure, for kinematics of the lumbar-pelvic region. Murphy
(2009) found ∆L5/S1 to be the most pertinent variable in prediction of lower back injuries in
rowing. Hence this variable was selected as the most appropriate dependent variable for this model, which represents changes in lumbar-pelvic motion during the loading phase of the rowing stroke.
The mean and standard deviations of the dependent variable, in addition to the independent variables which were extracted by the Stepwise regression model are shown in Table 5. Three predictors were significantly correlated with average resultant force; ASI vertical force at MHF, ASI average resultant foot force and ASI timing of heels down. With these variables the model achieved an R
2
and adjusted R
2
value of 0.43 and 0.41 respectively.
The relationship between foot force asymmetries and pelvic twist
12

Asymmetries in foot force application may cause kinematic alterations further along the kinematic chain, and this may possibly be characterized by rotation of the pelvis in the axial plane. Thus, pelvic twist was used as a dependent variable to see if asymmetries at the feet translate to a rotation of the pelvis.
The mean and standard deviations of the dependent variable, in addition to the independent variables which were extracted by the Stepwise regression model are shown in Table 6. Three predictors were significantly correlated with average resultant force; ASI vertical force at finish, ASI average resultant foot force and ASI timing of heels down. With these variables the model achieved an R 2 and adjusted R 2 value of 0.32 and 0.29 respectively.
Discussion
Rowing is a sport in which the parameters of injury, performance and training are all intimately related and in dynamic equilibrium. Therefore the purpose of this study was to examine aspects of technique which optimise resultant and horizontal foot force, in addition to investigating the effect of loading asymmetries on trunk and pelvis kinematics.
The first aim investigated discrete aspects of technique which strongly influenced average resultant forces applied to the foot-stretchers. The multiple regression model found that almost
50% of variance in foot force at MHF could be explained by sagittal plane hip and ankle kinematics. However, nearly 35% of variance in peak resultant foot force could be explained by kinematics of the hip alone. As such, in terms of producing large magnitudes of resultant foot force, hip kinematics appears to have a greater influence on foot force compared to knee or ankle kinematics. This idea is reflected in other activities with similar lower limb ROM, such as flywheel resisted squats and chair squats, where significantly greater contribution of the hip
13

joints have been demonstrated, over ankle and knee joint moments (Chiu and Salem, 2006,
Flanagan et al., 2003). Greater hip ROM during the stroke and greater degrees of hip flexion at
both the catch and MHF are suggested as key variables which contribute to greater resultant foot forces. As such, larger hip ROM seems to have been achieved in this data set through greater compression of the hips during the critical drive phase.
Additional regression analyses from this data set showed that high levels of hip flexion are associated with a smaller L5/S1 angle at the catch (R 2 =0.54). Both these kinematic
characteristics are considered beneficial in terms of performance and injury (Murphy, 2009).
This is because hip flexibility at the catch enables the rower to gain length without having to compensate through flexion of the lower back to gain additional stroke length. Furthermore, greater degrees of hip flexion at MHF suggests that the hips, and consequently the trunk, do not posteriorly rotate too early and that the legs remain engaged during the critical drive phase
(McArthur, 1997). The ability to apply high forces is dependent on maintaining postural position.
Therefore, by opening up the body too early by posteriorly rotating the trunk before the legs have reached full extension, mechanical advantage is lost and greater reliance is placed on body mass rather than the leg drive to continue increasing velocity of the boat. Therefore, hip flexibility is important in rowing as it enables the rower to get into an optimal position at the catch, in order to
achieve a long and powerful stroke (Kaehler, 2011).
Hip kinematics explains the greatest proportion of variance in generating large resultant foot force over the drive phase. However, locomotive studies have shown that it is mainly propulsive
ground reaction forces which contribute to large horizontal accelerations (Roberts and Scales,
2002, Hunter et al., 2005). In rowing, the goal is to achieve a large mean velocity of the boat in
the horizontal direction. However, the rowers’ feet are constrained at the metatarsals, and the
14

foot-stretchers against which forces are applied are orientated at 42°, thus making it difficult to apply purely horizontal forces. Consequently, in order to maximize the horizontal force ratio, technical application of foot force becomes very important. Investigations of individual force components have been done in cycling, where the effectiveness of force application has been
defined as the ratio of propulsive force to the total force applied onto the pedal (Davis and Hull,
1981, Ericson and Nisell, 1988). Similarly, in sprinting a runner is considered to have good
force application technique by exhibiting a high force ratio which favours the horizontal
component (Morin et al., 2011). For a given magnitude of resultant force output, different
strategies of force application which result in different RF may be used, thus increasing both the horizontal component of force and net forward acceleration. Given the direction in which the boat moves along the water, the same is true in the case of rowing. As such, it was necessary to consider the importance of foot force orientation, rather than simply the resultant force. The predictive kinematic variables which explained the greatest proportion of variance in RF at MHF were MHF knee flexion and ΔL5/S1. Examination of the relationship between knee angle at
MHF and RF show that the smaller the knee flexion at MHF the greater their RF (Table 4). This means that a more rapid extension of the knees encourages a greater horizontal force output at the foot-stretchers, thus improving foot force application technique. However, knee kinematics were responsible for just 25% of the variance in RF at MHF. When a stable L5/S1 is maintained from catch to MHF, an additional 15% of variance in RF can be explained. Exhibiting as little flexion of L5/S1 as possible at the catch, and maintaining a strong posture through the critical drive were two key variables which have previously been associated with improved performance
(Murphy, 2009). By keeping a strong connection between the lumbar spine and pelvis during the
drive phase, the lumbar segment will be prevented from rotating anteriorly beyond the pelvis
15

during the critical drive, thus putting the rower in a much stronger position to generate power
(Strahan et al., 2011). Consequently, a strong posture and rapid extension of the knees during
the critical drive phase are two key kinematic characteristics that explain some of the variance in
force application technique in elite rowers. This notion is supported by Jones and Miller (2002)
who stated that vertical movement in rowing is caused by flexion of the legs and swing of the upper body. Furthermore, keeping the lumbar spine stabilised and aligned with the pelvis as the trunk posteriorly rotates is an important technical feature of the rowing stroke which enables
resultant force at the handle to be optimised (McGregor et al., 2004). This is because the trunk
acts as an essential connection between the lower and upper limbs, thus enabling a more efficient
transfer of force to the handles and thus greater resultant force at the oars (Baudouin and
Hawkins, 2002).
Although three dimensional kinematics were derived and entered as inputs to the multiple regression model, the kinematic variables which had the greatest influence on resultant foot force and RF were those in the sagittal plane. This supports previous research which has shown outof-plane rotations to have low impact on technique, with only 0.53% of all out-of-plane rotations demonstrating significant changes with respect to rowing intensity and longitudinal training
(Murphy, 2009).
Further aims of this study were to examine the impact of asymmetries on variables of rowing technique which are thought to inflate the risk of injury. Previous studies have shown evidence
of asymmetries at the foot-stretchers (Baca et al., 2006). However, their influence on subsequent
aspects of rowing technique, namely; pelvic twist and L5/S1 flexion during the critical drive and performance has never been investigated.
16

The multiple regression model showed that over 40% of variance in ∆L5/S1 could be accounted for by various aspects of foot force asymmetry. Specifically, these were ASI average resultant force and ASI vertical force at MHF. ASI average resultant force was responsible for explaining the greatest proportion of variance in ∆L5/S1. This variable quantifies asymmetry of resultant foot force over the entire drive phase, thus is a key measure of asymmetry and may be useful in predicting kinematics of the lumbar-pelvic region during the rowing stroke. In addition to the magnitude of forces produced, the synchronicity in timing the leg drive was also highlighted as a possible contributor to ∆L5/S1. Timing differences in engaging the heels during the drive may result in kinematic asymmetries at the hips and knees. Asymmetries of the lower limbs have
previously been noted (Janshen et al., 2009), with discrepancies at the hip proving to be a
moderate predictor of L5/S1 flexion at the catch and MHF (Buckeridge et al., 2012). Whilst
little research has investigated asymmetries and their impact on the lumbar-pelvic region in rowing, clinical studies have shown that bilateral leg power asymmetries affect trunk stability
(Chung et al., 2008). Interestingly, the asymmetry variables which explained 31% of the
variance in pelvic twist were similar to the explanatory variables of ∆L5/S1 i.e. ASI average resultant force and ASI timing of heels down. Asymmetrical lifting has been shown to cause asymmetrical torques at the lower back, which resulted in increased spinal loading through
stabilizing co-contractions of the lumbar spine (Kingma et al., 1998). However, Plamondon
(1995) did not find 3D moments at L5/S1 that significantly differed between asymmetrical and
symmetrical lifting. They suggested that subjects may have minimized lateral flexion and torsional moments through compensatory twisting of their pelvis and lower limbs. Even in seated activities (such as rowing) pelvic asymmetry has been shown to cause a significant
increase in motion of the lumbar and thoracic spine (Al-Eisa et al., 2006). Biomechanically,
17

there is a clear influence of pelvic twisting on spinal kinematics, and in this study it may have originated from asymmetries in foot force production. Consequently, foot force asymmetries may have a small role in influencing both pelvic twist and ∆L5/S1. As such, measures of asymmetry at the footplates provide information regarding their influence on kinematics further along the chain of motion.
Measuring foot force in conjunction with rower kinematics has established many aspects of the rowing stroke, including force application and movement patterns, to be inherently dependent on one another. Hip kinematics were important in generating large resultant foot forces. However, there was a greater emphasis on a rapid drive through the knees and maintaining a strong lumbarpelvic posture when aiming to maximize the horizontal component of foot force. In terms of the impact of foot force asymmetries on variables that may influence injury risk, there was a moderate effect of asymmetries on ΔL5/S1 and pelvic twist during the critical drive. From this study, there is evidence to suggest that small changes in rowing technique have a notable influence on force production and asymmetries, which may ultimately influence rowing performance and injury risk. Additionally, interactions between force production and movement strategies provide a great amount of insight into aspects of technique that can be modified and improved to optimise force output and performance Having identifying discrete aspects of technique which improve foot force, and by quantifying the influence of asymmetries on joint and segment kinematics, it will be possible to provide an evidence based guide to coaches and athletes which aids performance optimisation and technique refinement.
18

AL-EISA, E., EGAN, D., DELUZIO, K. & WASSERSUG, R. 2006. Effects of pelvic asymmetry and low back pain on trunk kinematics during sitting: a comparison with standing. Spine (Phila Pa 1976), 31 , E135-43.
BACA, A., KORNFEIND, P. & HELLER, M. Year. Comparison of foot-stretcher force profiles between on-water and ergometer rowing. In: XXIV International Symposium on
Biomechanics in Sports, July 2006 Slazburg, Austria.
BAUDOUIN, A. & HAWKINS, D. 2002. A biomechanical review of factors affecting rowing performance. British Journal of Sports Medicine, 36 , 396-402.
BUCKERIDGE, E., HISLOP, S., BULL, A. & MCGREGOR, A. 2012. Kinematic asymmetries of the lower limbs during ergometer rowing. Medicine and Science in Sports and
Exercise, 44 , 2147-53.
BUCKERIDGE, E., BULL, A. & MCGREGOR, A. 2013. Foot force production and asymmetries in elite rowers. Sports Biomechanics (in press).
BULL, A. M. J., BERKSHIRE, F. H. & AMIS, A. A. 1998. Accuracy of an electromagnetic measurement device and application to the measurement and description of knee joint motion. Proceedings of the Institution of Mechanical Engineers Part H-Journal of
Engineering in Medicine, 212 , 347-355.
BULL, A. M. J., HOLT, B., WRAGG, P. & MCGREGOR, A. H. 2004. Validation of the use of a skin-mounted device to measure out-of-plane rotations of the spine for a rowing activity.
Journal of Musculoskeletal Research, 8 , 129 –132.
BULL, A. M. J. & MCGREGOR, A. H. 2000. Measuring spinal motion in rowers: the use of an electromagnetic device. Clinical Biomechanics, 15 , 772-776.
CAPLAN, N. & GARDNER, T. 2010. The influence of stretcher height on posture in ergometer rowing. European Journal of Sport Science, 10 , 263-268.
CAPLAN, N. & GARDNER, T. N. 2005. The influence of stretcher height on the mechanical effectiveness of rowing. Journal of Applied Biomechanics, 21 , 286-96.
CHIU, L. Z. F. & SALEM, G. J. 2006. Comparison of joint kinetics during free weight and flywheel resistance exercise. Journal of Strength and Conditioning Research, 20 , 555-
562.
CHUNG, L. H., REMELIUS, J. G., VAN EMMERIK, R. E. A. & KENT-BRAUN, J. A. 2008.
Leg Power Asymmetry and Postural Control in Women with Multiple Sclerosis.
Medicine and Science in Sports and Exercise, 40 , 1717-1724.
DAVIS, R. R. & HULL, M. L. 1981. Measurement of Pedal Loading in Bicycling. Journal of
Biomechanics, 14 , 496-496.
19

ERICSON, M. O. & NISELL, R. 1988. Efficiency of Pedal Forces during Ergometer Cycling.
International Journal of Sports Medicine, 9 , 118-122.
FIELD, A. 2009. Discovering Statistics Using SPSS, London, Sage.
FLANAGAN, S., SALEM, G. J., WANG, M. Y., SANKER, S. E. & GREENDALE, G. A. 2003.
Squatting exercises in older adults: Kinematic and kinetic comparisons. Medicine and
Science in Sports and Exercise, 35 , 635-643.
GROOD, E. S. & SUNTAY, W. J. 1983. A joint coordinate system for the clinical description of three-dimensional motions: application to the knee. Journal of Biomechanical
Engineering, 105 , 136-144.
HALLIDAY, S. E., ZAVATSKY, A. B. & HASE, K. 2004. Can functional electric stimulationassisted rowing reproduce a race-winning rowing stroke? Archives of Physical Medicine and Rehabilitation, 85 , 1265-1272.
HASE, K., KAYA, M., ZAVATSKY, A. B. & HALLIDAY, S. E. 2004. Musculoskeletal loads in ergometer rowing. Journal of Applied Biomechanics, 20 , 317-323.
HOFMIJSTER, M. J., SOEST, A. J. & VAN KONING, J. J. D. 2008. Rowing skill affects power loss on a modified rowing ergometer. Medicine and science in Sport and Exercise, 40 ,
1101-1110.
HOLT, P. J., BULL, A. M., CASHMAN, P. M. & MCGREGOR, A. H. 2003. Kinematics of spinal motion during prolonged rowing. International Journal of Sports Medicine, 24 ,
597-602.
HUNTER, J. P., MARSHALL, R. N. & MCNAIR, P. 2005. Relationships between ground reaction force impulse and kinematics of sprint-running acceleration. Journal of Applied
Biomechanics, 21 , 31-43.
JANSHEN, L., MATTES, K. & TIDOW, G. 2009. Muscular coordination of the lower extremities of oarsmen during ergometer rowing. Journal of Applied Biomechanics, 25 ,
156-164.
JONES, C. & MILLER, C. 2002. The mechanics and biomechanics of rowing. [Online].
Available: http://www.yorkshirerowing.co.uk/Files/Coaching%20+%20Training/Rowing%20Mecha nics%20+%20Biomechanics.pdf
[Accessed].
KAEHLER, B. 2011. Coach Kaehler: Ease Into The Catch. [Online]. Available: http://www.row2k.com/columns/index.cfm?action=read&ID=520&site=features
[Accessed 20/11/12].
KINGMA, I., VAN DIEEN, J. H., DE LOOZE, M., TOUSSAINT, H. M., DOLAN, P. &
BATEN, C. T. M. 1998. Asymmetric low back loading in asymmetric lifting movements is not prevented by pelvic twist. Journal of Biomechanics, 31 , 527-534.
20

LOH, J. H. H., BULL, A. M. J., MCGREGOR, A. H. & SCHROTER, R. C. Year.
Instrumentation of a Concept II rowing ergometer for kinetic & kinematic data acquisition. In: The Engineering of Sport 5, 2004. Sheffield, 173-179.
MCARTHUR, J. 1997. High Performance Rowing, Marlborough., The Crowood Press Ltd.
MCGREGOR, A. H., BULL, A. M. J. & BYNG-MADDICK, R. 2004. A comparison of rowing technique at different stroke rates: A description of sequencing, force production and kinematics. International Journal of Sports Medicine, 25 , 465-470.
MCGREGOR, A. H., PATANKAR, Z. S. & BULL, A. M. J. 2005. Spinal kinematics in elite oarswomen during a routine physiological "step test". Medicine and Science in Sports and Exercise, 37 , 1014-1020.
MCGREGOR, A. H., PATANKAR, Z. S. & BULL, A. M. J. 2008. Do men and women row differently? A spinal kinematic and force perspective. Proceedings of the Institution of
Mechanical Engineers Part P-Journal of Sports Engineering and Technology, 222 , 77-83.
MCNEELY, E. & ROYLE, M. 2002. Skillful Rowing, UK, Meyer & Meyer Sport.
MORIN, J. B., EDOUARD, P. & SAMOZINO, P. 2011. Technical Ability of Force Application as a Determinant Factor of Sprint Performance. Medicine and Science in Sports and
Exercise, 43 , 1680-1688.
MURPHY, A. J. 2009. Elite Rowing: Technique and Performance.
PhD, Imperial College
London.
PLAMONDON, A., GAGNON, M. & GRAVEL, D. 1995. Moments at the L(5)/S-1 Joint during
Asymmetrical Lifting - Effects of Different Load Trajectories and Initial Load Positions.
Clinical Biomechanics, 10 , 128-136.
ROBERTS, T. J. & SCALES, J. A. 2002. Mechanical power output during running accelerations in wild turkeys. Journal of Experimental Biology, 205 , 1485-1494.
ROBINSON, R. O., HERZOG, W. & NIGG, B. M. 1987. Use of force platform variables to quantify the effects of chiropractic manipulation on gait symmetry. Journal of
Manipulative and Physiological Therapeutics, 10 , 172-176.
STRAHAN, A. D., BURNETT, A. F., CANEIRO, J. P., DOYLE, M. M., O'SULLIVAN, P. B.
& GOODMAN, C. 2011. Differences in Spinopelvic Kinematics in Sweep and Scull
Ergometer Rowing. Clinical Journal of Sport Medicine, 21 , 330-336.
21

Figure 1: Figure depicting the locations of FOB sensor placements on a rower.
Table 1: Anatomical landmarks digitized during static recordings.
Anatomical Landmark
Right posterior superior iliac spine (RPSIS)
Left posterior superior iliac spine (LPSIS)
Right anterior superior iliac spine (RASIS)
Left anterior superior iliac spine (LASIS)
Lateral femoral epicondyle (LEPI)
Medial femoral epicondyle (MEPI)
Distal apex of the lateral malleolus (LMAL)
Distal apex of the medial malleolus (MMAL)
Dorsal aspect of the fifth metatarsal head (MET5)
Hip joint centre (HJC)
S2
S2
S4
S4
Stored as vector offset from:
S2
S2
S4
S4
Global origin
S2
22

Table 2: Explanatory inputs into Stepwise Regression models.
Kinematic Explanatory
Variables
ASI Explanatory Variables
Catch 3D ankle angles
MHF 3D ankle angles
Finish 3D ankle angles
Catch 3D knee angles
MHF 3D knee angles
Finish 3D knee angles
Catch 3D hip angles
MHF 3D hip angles
Finish 3D hip angles
ASI average resultant force
ASI resultant force at catch
ASI resultant force at MHF
ASI resultant force at finish
ASI average vertical force
ASI vertical force at catch
ASI vertical force at MHF
ASI vertical force at finish
ASI average horizontal force
Catch 3D L5/S1 angles
MHF 3D L5/S1 angles
ASI horizontal force at catch
ASI horizontal force at MHF
Finish 3D L5/S1 angles ASI horizontal force at finish
Ankle sagittal plane ROM ASI timing of peak vertical force
Knee sagittal plane ROM ASI timing of peak horizontal force
Hip sagittal plane ROM ASI timing of heels down
ASI vertical force at finish
Table 3: Descriptive statistics of the dependent variable (Average resultant foot force) and predictor variables, in addition to P values and regression coefficients of the predictor variables.
Dependent
Predictor
Predictor
Predictor
Variable
Average Resultant force
MHF hip flexion
Hip ROM
Catch ankle flexion
Mean Standard
Deviation
8.99
N/kg
102.15°
103.00°
75.90°
1.12 N/kg
9.40°
12.30°
10.61°
P value
-
0.00
0.00
0.02
Regression
Coefficient
0.02
0.02
0.02
23

Table 4: Descriptive statistics of the dependent variable (RF at MHF) and predictor variables, in addition to P values and regression coefficients of the predictor variables.
Variable Mean
Dependent
Predictor
Predictor
RF at MHF
ΔL5/S1
MHF knee flexion
0.81
1.16°
75.88°
Standard
Deviation
0.07
2.97°
9.55°
P value
-
0.00
0.00
Regression
Coefficient
-0.009
-0.004
Table 5: Descriptive statistics of the dependent variable (ΔL5/S1) and predictor variables, in addition to P values and regression coefficients of the predictor variables.
Variable Mean Standard
Deviation
1.16° 2.97° Dependent
ΔL5/S1
Predictor ASI average resultant force
6.26%
Predictor ASI vertical force at MHF 24.25%
Predictor ASI timing of heels down 22.44%
4.60%
15.90%
15.34%
P value
-
0.00
0.00
0.00
Regression
Coefficient
-
0.060
0.017
0.082
Table 6: Descriptive statistics of the dependent variable (Pelvic Twist) and predictor variables, in addition to P values and regression coefficients of the predictor variables.
Variable Mean Standard
Deviation
2.56° 3.55° Dependent
Predictor
Pelvic twist
Predictor ASI timing of heels down 22.44%
Predictor ASI vertical force at finish 16.65%
ASI average resultant force
6.26%
15.34%
12.98%
4.60%
P value
-
0.00
0.02
0.01
Regression
Coefficient
-
0.082
0.001
0.024
24