Saddle and leg forces during lateral movements in dressage



Reasons for performing study: In the equestrian world it is assumed that riders use changes in weight distribution and leg forces as important instruments to give horses directions about speed and direction of movement. However, the changes of these forces have never been quantified.

Objectives: To investigate the distribution of normal forces (perpendicular to surface) underneath the saddle and of normal forces exerted by the rider's legs during lateral movements.

Materials and methods: Eleven riders performed 3 different exercises: riding straight ahead, shoulder-in and travers at trot. Three saddle force systems were used simultaneously. The magnitudes of the forces were summed for the total area, the inside and the outside half of the saddle and inside and outside leg. Mean and maximum summed forces were analysed statistically.

Results: The saddle forces showed a rhythmic pattern but leg forces were more irregular. Mean total saddle force was lower (P = 0.006) when riding straight ahead (671 ± 143 N) than when riding shoulder-in (707 ± 150 N) or travers (726 ± 165 N). Mean inside saddle force was higher (P = 0.003) when riding travers (468 ± 151 N) than when riding straight on (425 ± 121 N) or shoulder-in (413 ± 136 N). Maximum outside leg force was higher (P = 0.013) when riding travers (47.2 ± 33.9 N) than when riding straight on (31.6 ± 24.1 N) or shoulder-in (34.2 ± 27.3 N).

Conclusion: The study helps to give a biomechanical background to well established but intuitive horse riding techniques.


In equestrian sports, the communication between rider and horse is crucial. In case of miscommunication, unsafe situations may result and even accidents may happen. Further, unclear aids of the rider can lead to mental stress in the horse and thereby might negatively influence its welfare.

The rider has several options to give aids to the horse in order to change either speed or direction of movement. One of these is the use of reins and the bit. Measurement devices to measure rein force have been developed (Clayton et al. 2005; Warren-Smith et al. 2005) and have been used to quantify this force during specific equitation movements (Warren-Smith et al. 2007) and to evaluate the effect of martingales and elastic rein inserts on horse behaviour and rein tension (Heleski et al. 2009). Changes in weight distribution of the rider and application of forces by the rider's legs are traditionally considered principal factors in correct horse riding (Decarpentry 1971). De Cocq et al. (2009) showed that a change in body position of the rider indeed results in a changed normal force distribution underneath the saddle, but it is not clear to what extent this change in force distribution is indeed used to instruct the horse when performing dressage exercises.

In lateral movements in dressage, such as shoulder-in and travers, the horse is required to proceed in an orientation where the long axis of its body is not aligned with the direction of movement. According to the riding theory, this difference between body orientation and horizontal motion is affected and maintained through a series of aids that include a shift in weight distribution, a change in the position of the rider's leg and the asymmetrical exertion of leg forces (Hölzel et al. 1992). This theory was tested quantitatively by measuring the distribution of normal (i.e. perpendicular to the surface) forces underneath the saddle and of normal forces exerted by the rider's legs during lateral movements of the horse.

Materials and methods

Horses and riders

Eleven horse-rider combinations were used during this study. All riders rode on their own horse and one rider rode 2 additional horses. There were 9 geldings, one mare and one stallion. Ten horses were Warmblood horses and one horse was a Fjord horse. The horses were 9.7 ± 2.5 years old, had a mass of 572.3 ± 53.0 kg and withers height of 1.70 ± 0.09 m. There were one male and 8 female riders. The riders were 24.3 ± 5.8 years old, with a body mass of 66.7 ± 8.0 kg and height of 1.73 ± 0.08 m. The horses were clinically sound and fit to perform. The combinations had a competition level in dressage that was at least intermediate or higher. The horses were equipped with their own tack. Riders were not allowed to use spurs.

Measurement equipment

Normal forces underneath the saddle and the rider leg forces were measured with 3 saddle force systems (Pliance)1. Each saddle force systems consisted of 2 halves with 128 sensors each. The sensors had a size of 25 x 37.5 mm. One system was placed underneath the saddle, with the other 2 on the right and left sides underneath the saddle flaps. The systems were connected via a connection plug and measured synchronically with a frequency of 26 Hz. Before the measurements, all 3 saddle systems were calibrated using an air pressured calibration ridge. The calibration was tested after each measurement day. If the variation coefficient of the sensors was below 10%, measurements were considered reliable (de Cocq et al. 2006).

Data collection

After calibration, the 3 saddle force systems were placed in a custom made saddle pad. The saddle pad was placed on the back of the horse. When saddling the horse, it was checked that the systems were placed symmetrically and that the front of the saddle was on top of the first row of normal force sensors. After the girth was tightened, the saddle systems were set at zero.

Normal force measurements were performed at sitting trot in 3 conditions: proceeding along a straight line, shoulder-in and travers (Fig 1). The order of the 3 conditions was randomised. Before the measurement the horses had a warming-up period of 10–15 min, which included 5 min of walk and 5–10 min of trot on both leads. Riders were allowed to choose their preferred lead for the measurements. For all 3 conditions, the left side of the horse was the inside at the left lead and the outside at the right lead. The right side of the horse was the inside at the right lead and the outside at the left lead. Infrared gates connected to a time registration system were used to control the speed. The infrared gates stood 11 m apart and a variation in trial duration of maximally 0.4 s was accepted. A minimum of 6 trials per condition in this speed range was measured. Three digital video cameras were used to film the trials. One camera was facing the front of the horse, one the hind of the horse and one viewed the horse from the side. The video shots were used for the evaluation of the exercises by independent judges afterwards.

Figure 1.

Body position of the horse and direction of movement during the 3 exercise conditions. (a) Riding straight ahead; (b) shoulder-in; (c) travers. The large arrows indicate the direction of movement of the horse; The ∩ indicate the position of the horse's limbs and the small arrows indicate the direction of movement of the horse's limbs (

Data analysis

The saddle force was calculated by summation of the magnitudes of the forces measured by the individual sensors of the saddle device that was placed underneath the saddle. Forces were calculated for the total surface and for the left and right half of the saddle device (de Cocq et al. 2009). The leg forces were calculated by summation of the magnitudes of the forces measured by specific areas of the saddle devices that had been placed on the left and the right side under the saddle flaps. These areas consisted of 11 rows by 8 columns of sensors (Fig 2a).

Figure 2.

Specified areas for the saddle and leg force measurements. (a) The grey areas indicate the sensors that were used to measure leg and saddle force; (b) the red areas indicate malfunctioning sensors that were excluded from the analysis.

As some riders performed the exercises on the left lead and others on the right, the measurements of the inside and the outside of the horse were grouped. The differences between inside and outside saddle and leg forces were determined by subtracting the inside forces from the outside force. The mean force of each trial was calculated and the force peaks of the data identified using a routine that marks a series of data points that are neighboured by at least 8 points with lower values. From these force peaks the mean peak force (‘maximum force’) was calculated. Furthermore, the data of the first trial of each horse rider combination of the total saddle surface, the inside and outside leg were imported into Matlab2 and then power spectrums of the signals were calculated using Fourier analysis. For the Fourier analyses all trials were sampled and the same number of time points (60) used. The frequency with the highest power was called the main fundamental frequency.


Each trial consisted of 5 or 6 strides. All trials were judged by 2 independent national dressage judges on a scale from 0–10. The trails were judged on lateral bending, the number of tracks used, rhythm and the head and neck position. Trials with a mark above a 5.5 were considered correctly performed exercises. Mean ± s.d. were calculated from all 6 trials and for all correctly performed trials and all incorrectly performed trials for each horse-rider combination at each condition. Data were checked for normality of distribution using a Kolmogorov-Smirnov test and analysed statistically in a GLM-repeated measures test followed by a post hoc Bonferroni test using SPSS software3. The condition (straight ahead, shoulder-in or travers) was the within-subject factor. The lead (right or left) was the between-subject factor. In case data were not distributed normally, the nonparametric Friedman test was used to test for differences between the conditions. If a significant difference was found with the Friedman test, a pairwise comparison was made using the Wilcoxon test. The differences between the left and right lead were tested using the Mann-Whitney test when data were not normally distributed. A P value of <0.05 was considered statistically significant.


Calibration tests

The calibration results at the end of the first and second measurement days were within the accepted range (variation coefficient <10%). The calibration results of the third day were, however, not acceptable. This problem was caused by malfunctioning sensors. These sensors were excluded from the data analysis for all horse-rider combinations (Fig 2b).

Saddle force and leg force

Six riders rode on the right lead and 5 on the left. The pattern of the leg forces was irregular, while the pattern of the saddle forces had a rhythmic, regular appearance, as shown in the typical examples in Figure 3. There were no significant differences, neither between riding on the right or left lead, nor between correctly and incorrectly performed exercises. Therefore, left and right, outside and inside data were pooled and the results of all trials are given.

Figure 3.

Typical examples of saddle and leg force patterns. (a) Typical example saddle force pattern; (b) typical example power spectrum saddle force; (c) typical example leg force pattern, with one force peaks during each step; (d) typical example power spectrum leg force, with one force peaks during each step; (e) typical example leg force pattern, with one force peak during each stride; (f) typical example power spectrum leg force, with one force peak during each stride; (g) typical example leg force pattern, with low base value with irregular signals; (h) typical example power spectrum leg force, with low base values with irregular signals.

Mean total saddle force was significantly lower (P = 0.006) when riding straight ahead (671 ± 143 N) than when riding shoulder-in (707 ± 150 N) or travers (726 ± 165 N). Mean inside saddle force was significantly higher (P = 0.003) when riding travers (468 ± 151 N) than when riding straight on (425 ± 121 N) or shoulder-in (413 ± 136 N). The difference in maximum saddle force between the outside and inside of the saddle was significantly higher (P = 0.038) when riding shoulder-in (75.0 ± 212.0) compared to riding straight on (38.3 ± 189.9) or travers (-3.4 ± 197.0). The main fundamental frequency of the total saddle force was 2.4–2.5 Hz and did not differ between the exercises (Table 1). Maximum outside leg force was significantly higher (P = 0.013) when riding travers (47.2 ± 33.9 N) than when riding straight on (31.6 ± 24.1 N) or shoulder-in (34.2 ± 27.3 N). The main frequencies of the leg forces were lower and more variable compared to the saddle force (Table 2).

Table 1. Saddle force while riding straight on, shoulder-in and travers (mean ± s.d.)
VariablesStraight onShoulder-inTraversP value
  1. Max is the mean peak force; Δ is the difference between the outside and the inside force (outside minus inside); *significantly different (P<0.05, Bonferonni correction).ab Values with different superscripts are significantly different.

Mean total saddle (N)671 ± 143b707 ± 150a726 ± 165a0.006*
Max total saddle (N)1325 ± 2471381 ± 3361341 ± 2510.913
Mean inside saddle (N)425 ± 121a413 ± 136a468 ± 151b0.003*
Max inside saddle (N)817 ± 213778 ± 213833 ± 2170.307
Mean outside saddle (N)453 ± 102469 ± 112474 ± 1150.761
Max outside saddle (N)856 ± 142853 ± 145829 ± 1540.913
Δ Mean saddle (N)28.1 ± 116.055.6 ± 135.75.9 ± 134.20.103
Δ Maximum saddle (N)38.3 ± 189.9a75.0 ± 212.0b−3.4 ± 197.0a0.038*
Main fundamental frequency (Hz)2.4 ± 0.12.5 ± 0.12.4 ± 0.20.337
Table 2. Rider leg force while riding straight on, shoulder-in and travers (mean ± s.d.)
VariablesStraight onShoulder-inTraversP value
  1. Max is the mean peak force; Δ is the difference between the outside and the inside force (outside minus inside). *Significantly different (P<0.05, Bonferonni correction).ab Values with the different superscripts are significantly different.

Mean inside leg (N)17.6 ± 10.021.4 ± 12.620.7 ± 12.70.529
Max inside leg (N)34.1 ± 20.342.1 ± 22.434.7 ± 18.20.529
Mean outside leg (N)16.8 ± 23.621.0 ± 25.026.3 ± 28.70.060
Max outside leg (N)31.6 ± 24.1a34.2 ± 27.3a47.2 ± 33.9b0.013*
Δ Mean leg (N)−0.8 ± 27.9−0.4 ± 29.35.6 ± 31.40.178
Δ Maximum leg (N)−2.4 ± 31.3−7.9 ± 36.312.5 ± 38.00.078
Main fundamental frequency inside leg (Hz)1.6 ± 0.81.8 ± 0.81.8 ± 0.70.697
Main fundamental frequency outside leg (Hz)1.8 ± 0.62.0 ± 1.61.4 ± 0.80.723


This study showed that there were significant differences in force distribution over the saddle (measured indirectly as normal forces) in the different exercises performed, as is the force paradigm in the equestrian literature. For lateral movements in general, it has been postulated that the rider should always shift weight in the direction of movement of the horse (Stodulka 2006). For the travers, in which the hind quarter is brought to the inside and the horse bends to the inside, the weight shift should go towards the inside of the horse. The present study agrees with this, as mean inside saddle force was significantly higher in the travers than in both other conditions and it was the only condition in which the mean difference between outside and inside force was negative. Shoulder-in is a more complicated topic. The riding literature states that it is the only exception to the rule, i.e. here more weight should be put on the inside where the opposite would be expected based on the general rule. It is said that an inward weight shift is necessary in this case to lower the inner hip, which would facilitate heavier loading of the inside hindlimb (Stodulka 2006). In the present study no such effect could be demonstrated. There was a slight tendency towards the opposite effect, as the outside-inside difference of the peak saddle force in the shoulder-in condition was significantly higher than in both other conditions.

There was a significant difference in mean total saddle force between riding straight and both the lateral gaits. This difference was unexpected as the mean vertical force of the rider remains the same in the 3 situations and there is no net vertical displacement of the riders. As the saddle system is curved on top of the horse's back, it does not only measure vertical forces, but also horizontal forces. It therefore seems that the riders are gripping onto the saddle using their thighs in both shoulder-in and travers. This would explain why the mean total saddle force was higher during these exercises.

The maximal outside leg force of the rider was higher in travers than during riding straight on or shoulder-in. During travers the hindquarters are set to the inside. Riders have to use their outside leg to support this position and to indicate the direction of movement (Hölzel et al. 1992). There were no other significant changes in leg forces. It should be noted that the variation in leg force magnitudes was high, with standard deviations approaching the means in some cases. The leg force patterns (Fig 3) show that not only the force, but also the peak frequency of leg forces was highly variable. This figure demonstrates that the patterns of leg force are also very variable. This may well represent different strategies for using leg force by the individual riders. In general, the following patterns can be distinguished: a signal with each moving diagonal, a signal once during a stride cycle and very low base value with irregular signals. As the signals of the legs can be given independently of the movement of the horse, they may be better recognisable for the horse than signals emanating from changes in weight distribution. It would be interesting to investigate whether this type of pattern could be used for the assessment of rider quality.

During the experiment, problems occurred with malfunctioning sensors. These were identified during the calibration and henceforth removed from the analysis. Malfunctioning sensors is a common problem in saddle pressure analysis (Werner et al. 2002) and may, in our case, have been caused by a combination of sweat and folding of the saddle force system near the girth. It is, therefore advisable to cover the saddle system with an extra water resistant cover and calibration tests seem imperative in this type of research to detect malfunctioning sensors. As removal of malfunctioning sensors results in lower force values, the malfunctioning sensors were removed from the whole data set in order to not influence the comparison between the 3 conditions.

It should be realised that the outcome of this study may have been influenced by the use of long-established horse-rider combinations that may have developed entirely or partly compensated asymmetries over time. Further, the riders were of a reasonable level, but not of international standard. Rider quality might also affect the outcome of saddle force measurements.


By applying state-of-the-art technology, this study was able to measure, in terms of forces, weight and leg aids given in 2 lateral exercises as performed in dressage, compared to the standard situation of progression over a straight line with the longitudinal axis of the horse aligned to the line of progression. It appeared that leg aids were in line with expectations based on the equestrian literature, although inter-individual variation between riders was high. Shifts in weight distribution (as indicated by changes in normal saddle forces) also agreed with the equestrian literature in travers, but not when riding shoulder-in. Further and more detailed investigations into the exercise practiced in dressage are needed to confirm or falsify the empirically based theory behind this fascinating and heavily disputed equestrian discipline.


The authors thank the riders and horse owners who have participated in the experiment. We give special thanks to Anne Mariken Duncker for her invaluable support during the data collection.

Manufacturers' addresses

1 Novel GmbH, Munich, Germany.

2 MathWorks, Natick, Massachusetts, USA.

3 SPSS Inc., Chicago, Illinois, USA.

Conflicts of interest

The authors have declared no potential conflicts.