Saddle pressure patterns of three different training saddles (normal tree, flexible tree, treeless) in Thoroughbred racehorses at trot and gallop




Reasons for performing study: To a large extent the success of a racehorse depends on effective and health preserving training methods. An important issue is the prevention of back pain. The influence of different types of training saddles (normal tree: SA, treeless: SB, flexible tree: SC) on the saddle pressure patterns in racehorses have not previously been investigated. It is commonly assumed that SA limits the motion of the back especially in the lower thoracic region during gallop.

Hypothesis: SA produces higher pressures in the caudal part of the saddle at trot (rising trot), canter and gallop (both in a jockey seat) compared to SB and SC.

Methods: Saddle pressures were measured in 8 racehorses ridden on a training track at trot (3.5 m/s), canter (6.4 m/s) and gallop (12.6 m/s). Each horse performed the protocol with each saddle. To analyse the pressure distribution over the horse's back the pressure picture was divided into thirds (TDfront, TDmid, TDhind). The stride-mean loaded areas, forces and mean and peak pressures were determined.

Results: At canter and gallop, all 3 saddles were mainly loaded in TDfront (>80% of the rider's weight), with a decreasing gradient to TDmid and TDhind (<3%), which was least pronounced in SC. At trot, the load was shifted towards TDmid and TDhind (10–15%, each). High peak pressures occurred in TDfront at canter and gallop and in TDhind at trot.

Conclusions: The type of tree had no influence on the pressure picture of the caudal third at gallop. The high peak pressures observed in TDhind at trot in all saddles may limit the activity of the horse's back, which is of particular importance since trot is an integral part of the daily work.


Many different factors contribute to the successful training of a racehorse. The main goal is to lead the athlete to its peak performance with an individually adapted training regime and preferably without risking health problems. Nevertheless, the horse can become exercise intolerant. The causes are not always easily to pinpoint but mainly affect the respiratory and/or the musculoskeletal system (Martin 2004).

A closer look at the locomotor system reveals that back pain in particular is one of the most troublesome findings in sport horses. Numerous publications are dedicated to its aetiology, diagnostics, therapy and management (Jeffcott 1979; Jeffcott and Weaver 1992; Haussler 1999; Marks 1999; Martin and Klide 1999; Ridgway and Harman 1999; Weaver et al. 1999). Important risk factors related to wear and tear that have to be considered for preserving the soundness of the horse's back, are associated with the designated use and type of training of the horse, but also to the skills of the rider (Schollhorn et al. 2006) and the type and fit of the saddle (Harman 1999).

Although a sound back is an essential prerequisite for the development of the hindquarter mechanics in order to gallop efficiently, only a few aspects addressing saddling have been subject of studies in racehorses. Bowers and Slocombe (1999, 2005) have investigated the effect of saddle girth materials and applied tensions on performance in racehorses. They observed a reduction of the respiratory (Bowers and Slocombe 1999) and general performance (Bowers and Slocombe 2005) with increasing girth tension.

Werner (2006) is the only author to address the problem of saddle fitting in Thoroughbred racehorses and relate it to clinical and radiographic findings. It could be demonstrated that training with poorly fitted saddles resulted in an uneven force distribution of the rider's weight with high peak pressure values.

Different types of training saddles are used in racehorses. There are training saddles built of a normal wooden tree similar to that used in English saddles. It is commonly believed that when the horse gallops, these saddles interfere with the movement of the back in the thoracolumbar region because of their rigidity. Treeless training saddles are intended to solve this alleged problem but they are known for their lack of stability. Recent efforts in optimising fit of training saddles promoted the development of saddles with a flexible synthetic tree, which should have the stability of a saddle with a wooden tree and at the same time allow the horse's back to move freely.

The aim of this study was to investigate the longitudinal pressure distribution of the 3 different saddle types at gallop as well as at rising trot.

Materials and methods

Horses and rider

Eight clinically sound Thoroughbred racehorses were used in this study; age range 2–7, mean 4 years; bwt range 441–577 kg (estimated with Carroll and Huntington 1988); mean ± s.d. withers height 1.64 ± 0.38 m. All animals lived in the same barn and were stabled individually. They were trained by the same trainer and close to race fitness.

In all experimental trials the horses were ridden by the same experienced jockey (B.R.) with a bodyweight of 60 kg. She was instructed to pay attention to a balanced seat and to consistency in velocities between the saddle trials of each horse.


Three saddles with 3 different construction concepts were tested: SA was a race saddle with a wooden tree, SB a treeless saddle and SC had a flexible synthetic tree. The dimensions of the 3 saddles were determined with a measuring tape and are listed in Table 1. Each horse performed the exercise protocol within the scope of the daily training session and, therefore, the different saddles were tested on different days and in a random order.

Table 1. Dimensions of the 3 saddles
ParameterSaddle ASaddle BSaddle C
  • a

    Weight includes the 2 saddle pads and the saddle pressure measuring system. Black, headplate and tree; grey, panel.

  • b

    b Distance between the front edge of the gullet and the rear edge of the third billet.

  • c

    c Distance between the rear edge of the third billet and the rearmost part of the saddle panels.

Picture of tree concept and saddle panelsaWooden tree
inline image
inline image
Flexible tree
inline image
Weight (kg)
Gullet length (cm)444450
Gullet width (cm)4.53.55
Front panels width (cm)171625
Rear panels width (cm)121111
D1b (cm)151717
D2c (cm)292733
Total size of mask (cm2)147215001650

Data acquisition

The exercise protocol was performed on an oval sand training track. Three straight and flat sections of the track, of at least 100 m in length were selected as testing sections. The beginning and end of each section was marked with flagpoles. During the exercise protocol, the testing sections were ridden consecutively in different gaits and without interruption. The first section was ridden in a rising trot on the left hand (TR), the second at canter (G1) and the third at gallop (G2), both on the right hand and with the rider in a jockey seat. Video analysis confirmed that G1 was performed at a 3-beat leaping gait with the leading hind and trailing front limbs moving synchronously and G2 at a 4-beat leaping gait with the footfalls of the diagonal being dissociated (Barrey 2001). As saddle pressure measures are sensitive to changes in velocity (Bogisch et al. 2008), speed within horse and gait step has to be controlled for repeated measures. Therefore, speed was measured with a G3 W.I.N.D. GPS Sensor1 and speed information was displayed and recorded continuously on a CS600 Polar watch2. Split times were taken with stop watches as the horse passed the flagpole at the start and end of each section.

Saddle pressures were measured with a Pliance-X System1. The pressure sensitive mat (Pliance MSA600)2 consisted of 2 separate parts each with 128 sensors in a 16 × 8 (longitudinal × transverse) array. Every sensor had a size of 3.75 × 2.5 cm (9.38 cm2). The 2 mat parts were placed symmetrically on each side of the horse's back and were linked with 2 Velcro fasteners. The mat was set to zero base line before saddling and tightening the girth. The distance between the hind end of the mat and the caudal aspect of the dorsal process of L6 was measured before and after the measurement to check for mat displacement.

A rectangular foam pad (thickness 0.8 cm) combined with a rectangular foam stuffed cotton pad (thickness 3 cm) was placed between the pressure sensitive mat and the respective saddle. The caudal edge of the cantle was aligned to the rearmost sensor row of the pressure sensitive mat.

At the beginning of each trial, the stop watches, the recording watch of the GPS system and the data logger of the saddle pressure measuring system were started simultaneously.

Data analysis

Based on the split times, the velocity and the saddle pressure data of each sector were extracted from the continuous recordings. Mean sector velocities were calculated with the Polar Pro Trainer Equine Edition software2. Pressure data were exported from Pliance-x Expert 11.3.51 and analysed in Matlab3. Each pressure data string was cut into strides, with the maximum positive slope of the total saddle force curve chosen as stride cutting criterion. At the rising trot, the maximum slope occurred while the rider sat down. At canter and gallop, the maximum increase of force occurred while the horses impacted with the diagonal limbs. Each stride was time-normalised to 0–100% stride duration. Successive strides of a record were averaged to a mean stride.

Because the study focused on the longitudinal pressure distribution under each saddle, the pressure picture was subdivided transversely into thirds (TDfront, TDmid, TDhind). The aim of this subdivision was to assess 3 functional areas, adapted to each saddle and taking into account the measured sizes of the saddle panels (Table 1). The length of the anterior third was predetermined by the distance between the front edge of the gullet and the rear edge of the third billet (D1 in Table 1). For the 2 posterior thirds, the remaining gullet length was divided into halves. Concerning the width of the thirds all the loaded sensors were included. Because the boundaries of the various thirds did not always match with the sensor array, the pressures at the borders of the thirds were linearly approximated to sub-sensor values, i.e. the sizes of the thirds were scaled to 1/10 of a sensor length.

For the respective track sections, the mean velocities (m/s) and mean stride durations (s) were determined. For each third as well as for the total contact area of the saddle, the loaded area (A), average pressure (P) and sum of the forces (F) of all loaded sensors were calculated for every time point of the normalised stride. Respective stride-mean values Amean, Pmean and Fmean were calculated for the entire stride. Additionally, the highest peak pressure (Ppeak) ever reached by one of the sensors during the entire stride was determined within the thirds. To visualise and compare the pressure distribution between the saddles, the pressure pictures of the saddle mat were averaged at the time points of maximal total force (Ftotmax).

Statistical analysis

Statistical analysis of the data of all variables was performed in SigmaStat 3.54. Differences between saddles (SA, SB, SC) and saddle thirds (front, mid, hind) were tested separately for each speed level (TR, G1, G2) with a 2-way ANOVA for repeated measures (RM).

In an additional 2-way RM-ANOVA with both, speed level and saddle thirds as repeated factors, the influence of velocity was evaluated for each saddle separately. Applied velocities and the stride durations within the 3 speed trials were compared between the saddles with one-way ANOVA.

If any of the ANOVA indicated significant influences of the investigated factors, post hoc multiple comparison tests were done (Holm-Sidak method). Experiment-wise overall significance level was set at α= 0.05.


In the course of the study, one horse could not be measured with SC due to acute onset of lameness. Because of intermittent bucking at G1 and G2, data of another horse could only be analysed for the trot. The displacement of the pressure sensitive mat was <1.5 cm, independent of saddle or horse. This value accounts for less than half of a sensor length and is therefore considered to have a minor effect.

Mean velocities and stride durations of the different saddle trials at the 3 speed levels are listed in Table 2. Discrete values of the stride-mean parameters Amean, Pmean and Fmean together with Ppeak are listed in Table 3 whereas the relative distribution of each parameter to TDfront, TDmid and TDhind is depicted as a bar graph in Figure 1. The ANOVA revealed a significant saddle-third interaction for all parameters. Unless otherwise stated all reported differences hereafter are significant.

Table 2. Mean ± s.d. velocity and stride duration of the different trials
GaitNumber of horsesSaddleVelocity (m/s)Stride duration (s)
  1. Three speed levels: trot (TR), canter (G1) and gallop (G2). SA, saddle with normal tree; SB, treeless saddle; SC saddle with flexible tree. No significant differences were observed between saddle trials within the same speed level (P<0.05).

TR8SA3.6 ± 0.20.69 ± 0.01
8SB3.5 ± 0.20.70 ± 0.02
7SC3.5 ± 0.20.70 ± 0.02
G17SA6.3 ± 0.40.56 ± 0.01
7SB6.3 ± 0.40.55 ± 0.01
6SC6.5 ± 0.40.56 ± 0.01
G27SA12.8 ± 1.70.45 ± 0.02
7SB12.2 ± 1.60.45 ± 0.03
6SC12.9 ± 0.80.45 ± 0.02
Table 3. Mean ± s.d. values for loaded area (Amean), mean pressure (Pmean), peak pressure (Ppeak) and mean force (Fmean) of the different saddles in the respective third, at trot (TR), canter (G1) and gallop (G2)
ParameterGaitSaddleTDfrontTDmidTDhindTotal value
  1. SA, saddle with normal tree; SB, treeless saddle; SC saddle with flexible tree. TDfront, TDmid, TDhind; front, mid and hind third, respectively. Different superscripts indicate significant (P<0.05) difference between saddles within the respective third of the saddles. Post hoc test results of the saddle-third interaction after 2-way RM-ANOVA for each gait separately.

Amean[cm2]TRSA625 ±36a179 ± 26a184 ± 28ab988 ± 61a
SB659 ± 48b185 ± 27a123 ± 12a968 ± 71a
SC635 ± 78a268 ± 30b191 ± 32b1094 ± 122a
G1SA644 ± 24a116 ± 33a28 ± 14a788 ± 40a
SB702 ± 25b77 ± 37a9 ± 13a789 ± 49a
SC666 ± 45ab233 ± 38b45 ± 32a944 ± 83b
G2SA627 ± 35a116 ± 33a47 ± 17ab790 ± 40a
SB694 ± 23b85 ± 32b11 ± 13a789 ± 39a
SC619 ± 44a248 ± 41c57 ± 30b925 ± 58b
Pmean[kPa]TRSA6.3 ± 0.6a4.5 ± 0.5a5.8 ± 0.6a5.8 ± 0.5a
SB5.1 ± 0.8b5.1 ± 0.6a6.1 ± 0.6a5.2 ± 0.5ab
SC4.9 ± 0.8b4.4 ± 0.8a6.3 ± 0.6a5.0 ± 0.5b
G1SA8.5 ± 0.6a3.2 ± 0.7a1.2 ± 0.4a7.5 ± 0.5a
SB7.4 ± 0.3b3.1 ± 0.6a1.0 ± 1.0a7.0 ± 0.3ab
SC7.5 ± 0.7ab4.0 ± 0.7a1.7 ± 0.4a6.3 ± 0.6b
G2SA10.1 ± 0.8a3.8 ± 0.8a2.0 ± 0.4ab8.8 ± 0.8a
SB9.1 ± 1.1ab4.0 ± 0.8a1.3 ± 0.7a8.5 ± 1.0a
SC8.7 ± 1.5b5.4 ± 1.0b3.0 ± 0.5b7.5 ± 1.1b
Ppeak[kPa]TRSA25.7 ± 5.2a19.9 ± 4.6a30.1 ± 6.5a 
SB26.9 ± 7.1a32.7 ± 6.5b32.6 ± 6.6a 
SC26.1 ± 4.7a23.2 ± 4.5a31.8 ± 5.3a 
G1SA33.7 ± 6.3a13.3 ± 3.8a3.9 ± 1.2a 
SB36.3 ± 5.8a15.8 ± 6.0ab1.7 ± 1.7a 
SC33.5 ± 8.6a20.0 ± 7.4b6.0 ± 1.7a 
G2SA37.7 ± 4.4a15.2 ± 2.5a6.4 ± 2.0ab 
SB39.0 ± 4.5a21.1 ± 5.1ab2.6 ± 1.7a 
SC37.0 ± 9.2a25.8 ± 9.9b12.2 ± 3.3b 
Fmean[N]TRSA392 ± 56a81 ± 11a105 ± 11a 
SB337 ± 67ab94 ± 11a75 ± 13a 
SC314 ± 82b118 ± 28a119 ± 18a 
G1SA548 ± 51a38 ± 17a4 ± 3a 
SB522 ± 18a26 ± 17a2 ± 3a 
SC498 ± 75a93 ± 25b9 ± 7a 
G2SA637 ± 68a46 ± 20a9 ± 3a 
SB635 ± 86a36 ± 20a2 ± 3a 
SC539 ± 124b135 ± 44b17 ± 10a 
Figure 1.

Relative distribution of the stride-mean loaded area (Amean), mean pressure (Pmean) and mean force (Fmean) within the thirds of the 3 saddles at trot (TR), canter (G1) and gallop (G2). SA, saddle with normal tree; SB, treeless saddle; SC saddle with flexible tree. TDfront, TDmid, TDhind; front, mid and hind third of the saddle, respectively. Data are given as group mean ± s.e. Different indices specify significant (P<0.05) difference between saddles within the same third (see legendTable 3).

Comparison of saddles as a whole

With regard to the saddles as a whole, the total loaded area (total Amean) of saddle SC was 17–20% larger than that of both other saddles at G1 and G2 whereas at trot a similar tendency was noted but not significant (P = 0.058). As a consequence, Pmean of SC was lower than those of SA and SB at G2 and lower than that of SA at G1 and TR. As expected, total Fmean within each gait did not differ between the 3 saddles.

Gait dependent differences

In all saddles the total Amean at G1 and G2 was approximately 80% of the total Amean at trot. Between G1 and G2 the total Amean did not change. Total Pmean increased with increasing speed (TR<G1<G2).

Comparison between the thirds within each saddle

The comparison of the values between the thirds in each individual saddle (the first aspect of saddle-third interaction) showed distinct differences between TDfront, TDmid and TDhind in all saddles.

The largest fraction of the total Amean was concentrated on TDfront, both at trot (58–68%), G1 and G2 (67–89%) (Fig 1). From front to back, Amean decreased gradually in SB and SC at all gaits and in SA in both canter and gallop (P<0.05), whereas at trot TDmid and TDhind had comparable areas (P>0.8). The caudal degradation of Amean was lowest in SC.

At trot, Pmean was distributed fairly evenly between the thirds in all saddles. Nevertheless, Pmean of SA was lower in TDmid; the ones of SB and SC were higher in TDhind than in the other thirds. Regarding Ppeak, the differences between the thirds were comparable to those of Pmean in SA and SC whereas in SB, Ppeak was higher in TDmid and TDhind than in TDfront. At G1 and G2, Pmean and Ppeak decreased gradually from front to back in all saddles with SC having the least pronounced decline.

At trot in all saddles, about 60% of the Fmean was concentrated in TDfront with the other thirds sharing the remaining 40% fairly equally.

At G1 and G2, the distribution of Fmean between thirds was similar to that of Pmean, with the difference that the gradation between TDfront and TDmid was more accentuated.

Comparison between the saddles regarding each third

The comparison of the values between the saddles regarding each third (the second aspect of the saddle-third interaction), showed differences as indicated in Table 3 and Figure 1. The most important findings are summarised below.

At all speed levels Amean of TDfront was the largest in SB, whereas SC had the largest area in TDmid. Concerning TDhind, Amean at trot and G2 was smaller in SB than in SC but the areas of SA and SC were not different.

At trot, Pmean did not differ between saddles within the thirds except for in SA which had, compared to the other saddles, higher Pmean in TDfront. At G2, Pmean of SC was the lowest in TDfront (only significant compared to SA), the highest in TDmid, and higher than SB but equal to SA in TDhind.

At trot, Ppeak only differed within TDmid between the saddles: SB had higher peaks than SA and SC. At G1 and G2, Ppeak in TDfront did not differ among saddles. In TDmid, Ppeak of SC was higher than that of SA.

In regard to the percentage of the total force distributed among the saddle thirds, Fmean (%) of SC in TDfront was lower than that of SA and SB and in TDmid higher than SA at all gaits (Fig 1).

Comparison of the pressure distribution at maximum load

Group-mean pressure pictures at the moment of Ftotmax of each saddle at trot (TR) and gallop (G2) are shown in Figure 2.

Figure 2.

Mean pressure pictures at the moment of Ftotmax of the 3 saddles at trot (TR) and gallop (G2). SA, saddle with normal tree; SB, treeless saddle; SC saddle with flexible tree. In each graph, top is cranial. G2 data are presented for the right lead gallop. Data of trials where the horse was at left lead gallop were mirrored mathematically for inclusion in the mean graph. At trot, SA showed the tendency to bridge. SB and SC showed high pressure values in TDmid and especially in TDhind. At gallop the total Amean of all saddles was concentrated at TDfront and showed a characteristic left-right Ppeak asymmetry. The slightly higher Ppeak values and forward shift of the centre of pressure on the right side was a result of the right lead gallop. During forelimb protraction the proximal end of the scapula moves backwards and produces resistance underneath the headplate (von Peinen et al. 2009).


The purpose of this study was to evaluate and compare 3 different training saddles commonly used in racehorses, based on their saddle pressure patterns at relevant gaits and speeds.

Differences related to the seat

The rider's position and balance is an important factor for the distribution of load on the horse's back. The rider who participated in this study was a professional jockey. She had a well balanced seat and was able to accomplish the protocols as expected. Consequently, both velocity and stride duration were remarkably consistent within the different trials for each speed level (Table 2). At canter and gallop, where the rider stood in the stirrups riding in the jockey seat, the highest percentage of loaded area was concentrated in TDfront and did not increase with increasing speed. Consequently, Pmean was also highest in TDfront and the most extreme Ppeak was observed there (Fig 1). At both speed levels and with all saddles, Ppeak in TDfront was close to or exceeded 34.5 kPa. Nyikos et al. (2005) also divided the saddle area transversely into thirds and determined this value to be the upper threshold in the cranial third associated with back pain. In the present study, the horses' backs were each time visually assessed after exercise while removing the pressure sensitive mat in order to get possible hints for saddle soreness (von Peinen et al. 2010). The majority of horses showed the typical dry spots underneath the headplate with each saddle type. Nevertheless the Ppeak threshold of 48.9 kPa, which was observed at canter by von Peinen et al. (2010) to correlate with the clinical findings of dry spots, was not reached.

Because at gallop up to 90% of the load was concentrated around the withers, special attention has to be drawn to saddle constructive key points such as headplate or stirrup bar, but also to the thickness of the underlying pad. A nonfitting headplate in combination with an unsuitable pad can raise the pressure underneath the saddle even more (Harman 2004). Depending on the material, saddle pads may dampen and level out a part of the forces acting on the horse's back (Kotschwar et al. 2010). In the present study, a foam pad combined with a foam stuffed cotton pad was used. This combination is usually applied in continental Europe and also increasingly in Great Britain. Its influence on the pressure values in the region of the withers is hardly ever under consideration.

At G2 in the TDhind, Ppeak was 6.4 kPa for SA, 2.6 kPa for SB and 12.2 kPa for SC. The pressures did not reach the critical value of 31 kPa set by Nyikos et al. (2005) for the caudal third. These findings indicated that, as soon as the rider was riding in the jockey seat, none of the saddles loaded the caudal thoracic region and therefore, did not limit the motion of the back in this area.

In contrast, the pressure distribution at the trot showed a completely different picture. As a direct consequence of the rider's seat at the rising trot, TDmid and above all TDhind contributed remarkably to the total Amean in all 3 saddles, which resulted in a distinct displacement of the rider's load from the TDfront to the 2 other thirds (Fig 2). How and to what extent this load redistribution is transmitted to the back depends on the saddle construction. The construction of the saddle determines the stirrup position whereas its position on the horse's back defines the deepest point of the seat. Both factors influence the rider's seat and balance. Stirrups placed too far forward, combined with short stirrup leathers place the rider in a typical chair-seat position. The centre of gravity is shifted backwards and the rear end of the saddle is excessively loaded (Harman 2004). This phenomenon could be seen in all of the 3 tested saddles in the present study. The pressure values for TDhind were around the threshold of 31 kPa, which was defined by Nyikos et al. (2005) as the upper tolerated pressure limit for the caudal part of the saddle.

Defensive reactions to palpation of the thoracolumbar area are a frequent clinical finding in performance horses, especially in racehorses. In the authors' experience, these findings often occur together with unwillingness in training. The horses are conspicuously nervous, which results in an elevated neck and an extended back. In this situation, a badly fitting saddle may aggravate the pain in the thoracolumbar area particularly at the trot. Racehorses are warmed-up and cooled down at trot. A considerable amount of time during training is spent in this gait and it therefore contributes to the optimal development of the horse's fitness.

Differences related to the saddle construction

The 3 training saddles differed considerably concerning the concept of the trees and shape of the panels (Table 1). SA most closely resembled a conventional English saddle regarding weight and shape. The seat was deeper than in the other 2 saddles and the panels were firmly stuffed. They differed considerably from the panels of SC, which were remarkably soft. The degree of cushioning of SB was considered as intermediate.

Although SC had the widest front panels, this did not explain the larger Amean of SC compared to SA and SB as these differences were found to be greatest in TDmid. This observation was explained by different gullet length of the saddle. SC was 6 cm longer than SA and SB and the panels therefore had a larger total surface. Another outstanding difference between SC and the remaining 2 saddles was its ability to distribute part of the pressure from TDfront towards TDmid and TDhind, which was visible at trot and gallop in both parameters Pmean and Fmean. Beside the larger size of the panels and thus the larger contact area, the concept of the flexible tree, which is better able to adapt to the curvature of the horse's back, could explain the more even mean pressure distribution of SC.

In contrast to SC, Amean, Pmean and Fmean of SB at gallop were almost restricted to TDfront. As this saddle had no rigid structure except for a headplate-like construction, only very limited load could be shifted to the caudal parts of the saddle. At rising trot, Ppeak was particularly high in TDmid because of the direct and local impact of the rider's weight.

Even though SA was used in daily training, the shape of its tree turned out to be inappropriate for the majority of the study horses. The radius of the tree was too large, i.e. the saddle was too straight compared to the horses' backs. The bridging phenomenon was observable at trot, where Fmean, Pmean and Ppeak in TDmid were lower than in TDhind. Bridging is considered to be the saddle fitting fault least tolerated by horses (Harman 2004; Nyikos et al. 2005). It might be speculated that an appropriately fitted saddle with a normal saddle tree would have delivered better results. As training saddles are rarely custom fitted and usually used for several horses, inappropriate saddle fit represents the reality. In the present study the flexible tree of SC seemed to adapt best to different shapes of the horses' backs and is therefore a valuable solution.


Saddles with trees do not exert critical pressure on the lower thoracic back at canter and gallop. Therefore it can be assumed that they do not have a negative influence on the movement of this region. High pressure values were observed in all 3 saddles in TDfront at gallop and in TDmid and TDhind at trot. They reached the threshold values that were related to back pain in other studies.

The treeless saddle turned out to concentrate the rider's weight to one specific region (depending on the rider's seat). The saddle with the normal tree was critical because of the tendency to cause bridging effects. The saddle with the flexible tree generally showed a shift of load towards TDmid and TDhind resulting in a more even load distribution. Although this saddle had the highest Pmean in TDhind at trot, Ppeak in TDhind was nearly the same for all saddles. This study recommends reconsidering the construction of training saddles used in racehorses.


This study was supported by the foundation ‘Stiftung Forschung für das Pferd’.

Conflicts of interest

The authors have declared no potential conflicts.

Manufacturers' addresses

1 Polar Electro Oy, Kempele, Finland.

2 Novel GmbH, Munich, Germany.

3 The Math Works Inc., Natick, Massachusetts, USA.

4 SPSS Inc, Chicago, Illinois, USA.