Does a Working Knowledge of Learning Theory Relate to Improved Horse Welfare and Rider Safety?

ABSTRACT

Training and riding directly effects horse welfare, highlighting the potential for training methods to improve both horse welfare and human safety. Learning theory is considered the most appropriate scientific foundation for horse training methods, yet equestrians’ knowledge of learning theory is reportedly low. The relationship between equestrians’ knowledge of learning theory terminology (LT) and horse welfare and rider safety was investigated to determine if rider knowledge of LT contributes to improved horse welfare and human safety. A sample of 394 Australian recreational and sport horse riders completed an online survey. Ridden horse welfare was assessed using validated husbandry and behavioral indicators. Rider safety was assessed by asking participants about ridden accidents and injuries. Rider knowledge of LT was assessed by asking participants to correctly identify scenarios that depicted three key operant conditioning terms: negative and positive reinforcement, and punishment. Only 24.6% of the sample correctly identified all three terms (the criteria for having a basic knowledge of LT), suggesting knowledge of LT has increased among equestrians but remains low. However, rider knowledge of LT was not significantly related to improved horse welfare or rider safety. It is uncertain why knowledge of LT remains so low among equestrians and why rider knowledge of LT did not translate to welfare and safety benefits as predicted. In facilitating the development of workable solutions to address the dual issues of poor horse welfare and human safety, we explore several possibilities, including a proposed new horse training framework that may enable learning theory to be leveraged more fully and deliver expected benefits.

KEYWORDS:

• Equestrian
• horse-related safety
• horse welfare
• human–animal interaction
• learning theory
• systems thinking

Horse riding, including equestrian sport, is a dangerous recreational activity (Ball et al., 2007; Pounder, 1984; van Gilder Cooke, 2012), with riders facing life-threatening and life-changing injuries and death (Kreisfeld & Harrison, 2020; National Coronial Information System, 2020). The dangers to ridden horses are at least as great, with high rates of unrecognized ridden lameness and back pain (Buckley et al., 2013; Greve & Dyson, 2014) and behavioral problems leading to premature destruction (wastage) (Ödberg & Bouissou, 1999). Poor horse welfare and poor rider safety are ongoing issues that diminish community acceptance of the horse industry (Gordon, 2001; Ingle, 2021), and there is evidence that these two issues are related, with improving horse welfare likely improving human safety (Luke, Smith, et al., 2022; Randle, 2016). The most recent update of the Five Domains Model (Mellor et al., 2020) for assessing animal welfare emphasizes the role of training and riding (as well as husbandry) as important determinants of horse welfare, highlighting the potential for improved horse training methods to improve horse welfare and (likely) rider safety.

Modern horse training practices are largely based on traditional practices that were primarily utilitarian, often harsh, and had little regard for horse welfare (Waran et al., 2002). Horse training mostly involves applying an unpleasant stimulus to the horse until the horse exhibits the desired response, such as applying pressure to the mouth via the bit by pulling on the reins to signal the horse to stop, and/or the application of an unpleasant stimulus to deter unwanted behavior, such as whipping a horse that refuses a jump (Waran et al., 2002). While these stimuli correspond to the operant conditioning principles of negative reinforcement and punishment, respectively, the literature suggests that most equestrians, including qualified equestrian coaches, have a limited understanding of the theory underpinning these principles (Brown & Connor, 2017; Warren-Smith & McGreevy, 2008). This lack of understanding of learning theory among equestrians means that overly severe stimuli are frequently applied to horses, leading to chronic (and arguably needless) exposure to painful stimuli (McLean & Christensen, 2017). Moreover, a lack of theoretical knowledge of learning theory likely increases the risk of inconsistent and poorly timed stimuli that can create confusion and impede learning (Baragli et al., 2015). Horses trained and ridden in this way have diminished welfare, and this is related to an increased risk of accidents and injuries to humans (Luke, Smith, et al., 2022). To encourage the development and implementation of training methods that deliver better horse welfare, equitation scientists have advocated for horse training techniques based on a better theoretical understanding and more ethical application of learning theory principles, such as operant and classical conditioning (ISES, 2018; McGreevy, 2007; McLean & Christensen, 2017).

Learning theory covers associative learning (classical and operant conditioning) and non-associative learning (habituation and sensitization), and these are among the most widely accepted scientific models of animal learning (Cooper, 1998; McLean & Christensen, 2017; Skinner, 1953). The principles of operant conditioning underpin modern horse training, even though they are poorly understood by most equestrians (Brown & Connor, 2017; Warren-Smith & McGreevy, 2008). When understood and applied correctly, learning theory provides equestrians with a powerful means of shaping horse behavior to achieve training goals using light and subtle stimuli (Cooper, 1998; Perone, 2003). Operant conditioning proposes that animals are more likely to repeat behaviors that lead to positive outcomes (reinforcement) and less likely to repeat behaviors that lead to unpleasant outcomes (punishment) (Kowalski, 2018). As such, operant conditioning can be leveraged to encourage desirable natural behaviors, as well as teach novel behaviors through reinforcement, and extinguish unwanted behaviors through punishment (Waran et al., 2002).

When investigating equestrians’ knowledge of learning theory, researchers have focused on the fundamentals of operant conditioning; namely, positive reinforcement, negative reinforcement, and positive punishment (punishment). This is most likely because these concepts align with many lay horse-training methods that emphasize the use of negative reinforcement and to a lesser extent punishment (German National Equestrian Federation, 1997). Negative reinforcement relies on the pressure/pain continuum, such that an unpleasant stimulus is applied to the horse and is removed as soon as the horse offers the desired behavior (McLean & Christensen, 2017). Removing the unpleasant stimulus rewards the horse for performing the desired behavior, thereby increasing the likelihood that the next time the same stimulus is applied to the horse, the horse will repeat the desired behavior. Skilled trainers use pressures that are just strong enough to elicit a response, with subtle pressures of minimal duration the goal of ethical and humane horse training (McGreevy, 2007). Punishment, also used by equestrians, has the opposite effect, reducing the likelihood of a behavior being repeated (McLean & Christensen, 2017). An example of punishment is hitting the horse with a whip after it has refused a jump to communicate to the horse that refusing jumps is not a desired behavior. However, punishment is a less desirable training technique than negative reinforcement because of its numerous side effects, such as negative associations with the punisher and learned helplessness (McGreevy & McLean, 2009; Mills, 1998). Like punishment, negative reinforcement relies on an aversive stimulus to motivate behavior. If such stimuli are applied poorly – for example, if excessive pressures are used or the trainer’s timing is poor and the stimuli are not removed immediately that the correct response is offered – then stimuli intended as negative reinforcers become punishers and will be accompanied by punishment’s negative side-effects. Positive reinforcement, like negative reinforcement, increases the likelihood of a response by rewarding the response with something the horse finds pleasant (such as a food treat) (Skinner, 1953).

Traditionally, positive reinforcement has not been advocated in horse training owing to concerns it may not provide the same degree of control as negative reinforcement (McGreevy, 2007) and anecdotal accounts that it encourages unwanted oral investigative behavior (mugging) in the horse (Hockenhull & Creighton, 2010). However, there is growing interest in the role of positive reinforcement in horse training that likely coincides with greater recognition of the influence of affective and arousal states in animal learning (Starling et al., 2013). Horses trained using positive reinforcement are more likely to have a positive affective state which in turn encourages behavioral flexibility, a state that is likely to facilitate learning (Merkies & Franzin, 2021; Starling et al., 2013). Conversely, horses trained predominantly with aversive stimuli are likely to have both short- and long-term negative affective states (Panksepp, 1998). Negative affective states can adversely affect the human–animal bond and contribute to reduced performance (Haverbeke et al., 2008), outcomes most equestrians would see as undesirable. Furthermore, training approaches that precipitate negative affective states are inconsistent with good horse welfare, as assessed using the Five Domains Model (Mellor et al., 2020). Blended reinforcement (where minimally aversive negative reinforcement is coupled with positive reinforcement) may be the most appropriate horse-training approach as it offers the training efficacy and practicality of negative reinforcement along with the affective benefits of positive reinforcement (Sankey et al., 2010; Warren-Smith & McGreevy, 2007).

How equestrians select and apply different types of reinforcement and punishment will have a profound effect on their horse in terms of their horse’s behavior, arousal, and affect, and ultimately their welfare. Some may argue that knowledge of the scientific terminology associated with learning theory is not required for the correct application of its principles, and this may be so. However, the high frequency of ridden horse hyperreactive behavior, such as bucking and rearing (also termed conflict behavior), high wastage rates (premature deaths) of young horses due to behavioral problems, and high numbers of equestrian injuries and deaths suggest the reinforcement and punishment choices made by current equestrians are delivering poor outcomes for horses and riders. The lack of a sound theoretical understanding of learning theory principles is likely a contributing factor to these poor outcomes.

A lack of theoretical knowledge has been identified as a barrier to improved training outcomes for dogs (Todd, 2018), and the same likely applies to horses (Warren-Smith & McGreevy, 2008). In the absence of sound theoretical knowledge giving riders insight into how their training choices effect their horse, it is unsurprising that most equestrians are unable to leverage learning theory principles in a manner that protects horse welfare and maximizes learning (Brown & Connor, 2017; Mills, 1998). Terms such as “pressure and release” are generally better understood than negative reinforcement (Brown & Connor, 2017); however, the problem with such terms is that while they are more accessible to equestrians, they are piecemeal and do not provide a comprehensive theoretical framework. So, when difficulties arise in training, the equestrian’s training toolbox is inadequate. Moreover, reducing learning theory to phrases such as “pressure and release” risks oversimplifying it to such an extent that it appears inflexible and unsophisticated. So, while it may not be necessary to learn the scientific terminology per se, it can be argued that for an equestrian to utilize learning theory in a way that maximizes horse welfare and learning, then knowledge of more than basic terminology is necessary (Brown & Connor, 2017; McLean & Christensen, 2017; Warren-Smith & McGreevy, 2008). Therefore, in this study, an equestrian’s knowledge of learning theory terminology will be used as a surrogate for their knowledge of learning theory.

Developing more ethical horse-training practices based on learning theory to improve ridden horse welfare has been advocated by equitation scientists for almost two decades (McGreevy et al., 2005). Despite this advocacy and the potential benefits of this approach for horses and humans, knowledge of learning theory appears to be low among equestrians (Brown & Connor, 2017; Warren-Smith & McGreevy, 2008). A UK study comparing professional and amateur equestrians showed 33.8% of professionals and 12.5% of amateurs could correctly define negative reinforcement (Brown & Connor, 2017). In Australia, a study of qualified equestrian coaches found only 11.9% could correctly explain negative reinforcement (arguably the most widely used horse-training principle) (Warren-Smith & McGreevy, 2008). In this study, we attempted to provide an up-to-date measure of learning theory knowledge among Australian equestrians, and we investigated if rider knowledge of learning theory is related to better horse welfare, and, in turn, better rider safety, as predicted in the literature (Brown & Connor, 2017; McLean & Christensen, 2017; Warren-Smith & McGreevy, 2008).

Methods

Ethical Approval

This research was conducted in accordance with the protocol submitted to the Human Ethics Committee of Central Queensland University (Approval number 0000022790). All participants provided informed consent to participate in this study.

Procedure

Australian recreational and sport horse riders aged  ≥18 years were recruited via Facebook with the goal of capturing a national cross-section of riders from a wide range of riding disciplines. The survey was promoted in various discipline-specific interest groups, including camp drafting, dressage, endurance, showing, show jumping, and trail riding, and via the Facebook pages of several equestrian organizations, including Equestrian Australia, Horse Riding Clubs Association of Victoria, and Pony Club Australia. The survey was available online for approximately 12 weeks, beginning in May 2021.

Survey

Our anonymous online survey, hosted on the Qualtrics platform, was designed to examine horse welfare, rider safety, and rider satisfaction. All questions except three were closed, with the majority using a five-point Likert scale. Questions covered rider demographics, horse care, horse behavior on the ground and under saddle, riding equipment, rider accidents and injuries, rider satisfaction, and rider knowledge of learning theory terminology (see online Supplemental File 1 for full details). Rider demographic information included age, sex, postcode, rider competence level, equestrian organization membership, and preferred riding discipline. This paper reports the findings related to rider knowledge of learning theory terminology.

Ridden Horse Welfare

A systems thinking framework underpinned the assessment of horse welfare in this study. Unlike traditional reductionist approaches to science that see the world as similar to a machine that can be broken down into its constituent parts, systems thinking sees the world as an irreducible system where the whole is more than the sum of its parts (Capra & Luisi, 2014). This approach to understanding and assessing horse welfare is consistent with modern conceptualizations of welfare, such as the Five Domains Model, which recognize that welfare is a dynamic process to which all aspects of the animal’s life contribute (Mellor et al., 2020). Appreciating that welfare is dynamic is particularly relevant to ridden horses as welfare status may vary widely between rest and when the horse is ridden. Notwithstanding the complexities of assessing ridden horse welfare, there is currently no validated ridden horse welfare assessment tool that scrutinizes the ridden aspects of the horse’s life, making it necessary to construct a measure. Attempting to encapsulate our global view of ridden horse welfare, survey items included husbandry, health, and behavior on the ground and under saddle. Horse welfare item selection was based on items meeting three criteria: validated in the published literature, objective, and rider self-assessment was possible (see Table 1 for the complete list of indicators and supporting literature). While the survey could not assess welfare to the extent that a validated scale or on-farm assessment might, it did allow the relative welfare of sample horses to be ranked, which was sufficient for the purposes of the study.

Table 1. Management practices, medical conditions, and behavioral/other signals used to assess ridden horse welfare (in alphabetical order) (adapted from Luke, Smith, et al., 2022).

Management factors	Medical conditions	Signals during handling	Signals during riding
Competition (Jones & McGreevy, 2010; König v. Borstel et al., 2017)	Back soreness (Buckley et al., 2013)	Aggression (AWIN, 2015; Fureix et al., 2010)	Bolting (Dyson et al., 2018; McLean & Christensen, 2017)
Feeding (Lesimple, 2020)	Lameness (AWIN, 2015; Lesimple, 2020)	Apathy (AWIN, 2015; Fureix et al., 2010)	Bucking (Dyson et al., 2018; McLean & Christensen, 2017)
Housing (Lesimple, 2020)	Laminitis (Buckley, 2009; Lesimple, 2020)	Avoiding humans (AWIN, 2015)	Difficulty stopping (Dyson et al., 2018)
Transportation (Jones & McGreevy, 2010; Padalino et al., 2018)	Ulcers (Lesimple, 2020)	Hoof handling (Mansmann et al., 2011)	Ears back during transitions (Dyson et al., 2018; Fureix et al., 2010)
		Stereotypy (AWIN, 2015; Lesimple, 2020)	Rearing (Dyson et al., 2018; McLean & Christensen, 2017)
		Undesirable behavior (handling) (AWIN, 2015)	Rushing (Dyson et al., 2018)
			Saddle slip (Greve & Dyson, 2014; König v. Borstel et al., 2017)
			Spooking (Dyson et al., 2018; McLean & Christensen, 2017)
			Undesirable behavior (ridden)^a (Hockenhull & Creighton, 2013)

^a Undesirable behavior (ridden) was defined, using the criteria from Hockenhull and Creighton (2013), as behaviors, including move off before asked when rider mounts, pull/lean on the bit, jog when asked to walk, resist slowing, spook, trip or stumble, canter on the wrong leg, rush over jumps, buck or pigroot, resist turning when asked, refuse to move forward when asked, stop at jumps, run out when jumping, rear, bolt.

Most of the horse welfare questions were five-point Likert scale questions, where low scores indicated poorer welfare and higher scores indicated better welfare. All questions except one, which related to horse behavior while trotting or cantering (participants could select “not applicable – I didn’t trot or canter”), required a definitive response. Where “not applicable” was selected, this response was given the highest score, indicating good welfare; however, only eight participants (< 2%) selected this. Missing data that contributed to the relative welfare score were imputed using the median score for that item; however, only 13 records (3%) had missing data. Raw scores from the horse welfare items were added to calculate a relative horse welfare score for each horse. Scores could range from 0 to 84, with higher scores reflecting better horse welfare.

Human Safety

Human safety was assessed by asking participants to report falls, near-miss falls, and injuries sustained in the previous 7 days (to aid recall) and in the previous 12 months (to capture potentially more serious but less frequent accidents and injuries). Accident and injury scores were added to create a composite score for each rider.

Rider Knowledge of Learning Theory Terminology

Rider knowledge of learning theory was assessed by presenting riders with three commonly encountered scenarios that depicted one of the three key operant conditioning terms (negative reinforcement, positive reinforcement, and positive punishment) and asking them to identify the term demonstrated in the scenario (see online Supplemental File 1). Riders were deemed to have a basic working knowledge of learning theory if they correctly identified all three operant conditioning terms.

Statistical Analysis

As data were non-normally distributed, nonparametric statistical tests were used. Spearman’s rank correlation was used for parameter-level comparisons. Mann-Whitney and Kruskal-Wallis tests were used for group comparisons of continuous and ordinal variables, and chi-square tests were used for group comparisons of nominal variables. The significance level was p < 0.05. All analyses were conducted using SPSS (Version 26).

Results

Rider Demographics and Knowledge of Learning Theory

A total of 427 people completed the survey; however, 33 were excluded because they did not complete all three learning-theory questions, resulting in a final sample of 394. Respondents were mostly female (94.4%), their average age was 44.3 years (SD 13.9), the majority rated themselves as either intermediate (48.9%) or advanced (37.9%) level riders, and 81% belonged to an equestrian organization.

About one-quarter of riders (24.6%, n = 97) correctly identified all three operant conditioning terms. The ability to accurately identify negative reinforcement was only slightly higher, with 28.4% (n = 112) of participants answering correctly, while 69.5% (n = 274) incorrectly identified negative reinforcement as positive reinforcement, and 2.1% (n = 8) were unsure. Punishment was correctly identified by 59.1% (n = 233) of the sample; however, over one-third (34.8%, n = 137) incorrectly labelled punishment as negative reinforcement, 3.3% (n = 13) thought it was positive reinforcement, and 2.8% (n = 11) were unsure. The majority (98.7%, n = 389) correctly identified positive reinforcement (Table 2).

Table 2. Equestrians’ knowledge of three key components of learning theory: negative reinforcement, positive reinforcement, and positive punishment.

Participant responses	n	%
Negative reinforcement scenario (n = 394)
Negative reinforcement	112	28.4
Positive reinforcement	274	69.5
Punishment	0	0
Not sure	8	2.1
Positive reinforcement scenario (n = 394)
Negative reinforcement	1	0.3
Positive reinforcement	389	98.7
Punishment	2	0.5
Not sure	2	0.5
Punishment scenario (n = 394)
Negative reinforcement	137	34.8
Positive reinforcement	13	3.3
Punishment	233	59.1
Not sure	11	2.8
Working knowledge of learning theory (n = 394)
All three learning theory terms correctly identified	97	24.6
One or more learning theory terms incorrect	297	75.4

Learning Theory and Relative Horse Welfare

Overall, the relative horse welfare scores were skewed toward the sample horses having good welfare: the median relative welfare score was 75.0, IQR 9.0. Most sample horses (95.3%, n = 407) lived outdoors, and 52.9% (n = 226) lived outdoors with one or more conspecifics. Most horses had continuous access to forage (78.0%, n = 333), 15.7% (n = 67) were fed mostly forage meals plus some concentrate, and 6.3% (n = 27) were fed mostly concentrate meals with some forage.

To compare riders with and without knowledge of learning theory, riders who correctly identified all three operant conditioning terms were allocated to the “Yes” group, and riders who did not correctly identify all three terms were allocated to the “No” group. There was no significant difference in median relative horse welfare score for the “Yes” group compared with the “No” group (Mdn “Yes” group = 76.0, IQR = 9.5, Mdn “No” group = 74.0, IQR = 9.0, H₍₁₎ = 2.0, p > 0.05). Rider knowledge of learning theory is likely most relevant to horse welfare while the horse is being trained and ridden. To determine if husbandry practices might have been a confounding influence on relative horse welfare in relation to rider knowledge of learning theory, a post-hoc analysis was undertaken where a relative horse welfare score was calculated for each horse that excluded husbandry practices. However, when the data were controlled for husbandry practices, no difference in relative horse welfare score was found between riders with and without knowledge of learning theory (Median relative welfare scores (excluding husbandry) were 57.0 IQR = 9.0 and 58.0 IQR = 10.0, for the “Yes” and the “No” groups, respectively, H₍₁₎ = 1.3, p > 0.05).

Learning Theory and Horse Behavior

Most (59%, n = 252) horses in this study had performed at least one hyperreactive behavior in the seven days prior to the study, with 50.8% spooking, 22% bucking, 4% rearing, and 4% bolting. Median occurrence of hyperreactive behavior did not differ between riders with (“Yes” group) or without knowledge of learning theory (“No” group) (H₍₁₎= 1.1, p > 0.05).

Learning Theory and Human Safety

Rider accidents and injuries did not differ for riders with knowledge of learning theory (Median accident and injury score = 2.0, IQR = 4.0) compared with riders without knowledge of learning theory (Median accident and injury score = 1.0, IQR = 3.0, H₍₁₎= 2.8, p > 0.05).

Additional analyses were undertaken comparing riders who correctly identified negative reinforcement with riders who did not and comparing riders who correctly identified punishment with riders who did not. No meaningful differences between groups on the measures of horse welfare, ridden horse hyperreactive behavior, or rider accidents and injuries were found (see online Supplemental File 2).

Learning Theory and Equestrian Organization Membership

Based on a chi-square test, riders’ knowledge of learning theory did not significantly differ between members and non-members of equestrian organizations (χ² = 0.54, p > 0.05, Figure 1).

Figure 1. Membership of equestrian organizations and the percentage of riders with knowledge of learning theory.

Learning Theory and Rider Competency

The proportions of riders with knowledge of learning theory did not differ by rider competency (H₍₁₎ = 0.38, p > 0.05, Figure 2).

Figure 2. Rider competency and the percentage of riders with knowledge of learning theory.

Further analyses were conducted to determine if any differences were present across rider competence levels and equestrian organization membership/non-membership in terms of correctly identifying the individual terms negative reinforcement and punishment. No differences were found (see online Supplemental File 2).

Discussion

Despite knowledge of learning theory among Australian riders being found to have increased in the last decade, it remains low. Less than 25% of riders surveyed were able to correctly identify three key operant conditioning terms; namely, positive reinforcement, negative reinforcement, and positive punishment. In 2008, Warren-Smith and McGreevy (2008) reported 11.9% of qualified Australian equestrian coaches could correctly identify negative reinforcement. In our study of the Australian recreational and sport horse community, it was 28.1%, demonstrating a sizeable increase. This increase is likely due to the efforts of equitation scientists in promoting operant conditioning, and in particular negative reinforcement, as the key foundation of ethical horse training (McLean & McGreevy, 2010). For example, equitation scientists have been successful in securing the addition of learning theory to Pony Club Australia’s syllabus of instruction (Pony Club Australia, 2020). Equestrians’ knowledge of learning theory was similar across all rider competency levels, from beginner riders through to professionals. This finding is contrary to Brown and Connor (2017), who found a greater percentage of professional riders had knowledge of learning theory compared with amateur riders. Moreover, there was no significant difference in knowledge of learning theory for members of equestrian organizations (25.4%) and non-members (21.3%).

Although knowledge of learning theory has increased among riders, rider knowledge of learning theory (as assessed in this study) did not translate into a welfare benefit for horses or a safety benefit for riders. This finding is inconsistent with what is predicted in the literature, which suggests improved rider knowledge of learning theory has the potential to improve horse welfare and rider safety (Brown & Connor, 2017; Hawson et al., 2010; Warren-Smith & McGreevy, 2008). Understanding the disconnect between rider knowledge of learning theory and horse welfare and rider safety, as found in this study, may offer valuable insights to help direct future research efforts.

The lack of relationship between rider knowledge of learning theory and horse welfare could be due to poor survey construction; for example, our ridden horse welfare measure might have lacked validity. However, to mitigate this risk, the Five Domains Model (Mellor et al., 2020) and validated horse welfare indicators (AWIN, 2015; Lesimple, 2020) were leveraged to develop the horse welfare assessment items. Equally, our learning theory items might have been poorly constructed; however, to minimize this risk they were modelled on items used in previous research (Warren-Smith & McGreevy, 2008).

Study design may also have contributed to the lack of relationship between knowledge of learning theory and horse welfare. To achieve optimal learning, reinforcement and punishment must be applied correctly, consistently, and with precise timing (McLean & Christensen, 2017; Mills, 1998), factors that are impossible to measure with a survey. The inability to assess participants’ application of learning theory was a limitation of this study that was impossible to avoid. However, the high incidence of hyperreactive behavior in ridden horses (Hockenhull & Creighton, 2013; Luke, Smith, et al., 2022), and the extensive use of harsh equipment utilized by equestrians to control horses (Hill et al., 2015; McGreevy et al., 2012; Weller et al., 2021), combined with repeated studies demonstrating rider knowledge of learning theory is low (Brown & Connor, 2017; Warren-Smith & McGreevy, 2008), suggest these results are likely an accurate reflection of the equestrian population, although replication of this study would be valuable.

Accepting these findings as accurate raises two important questions: why has knowledge of learning theory remained so low among equestrians and why did rider knowledge of learning theory not translate to horse welfare and rider safety benefits as predicted? More research is needed, and to help guide future research, we explore in detail these two complex questions.

One plausible explanation for the low number of equestrians with knowledge of learning theory is that equestrians may feel there is no problem to be solved, so there is no need to change existing training methods. There is evidence for this from the racing industry, where racing insiders reject horse welfare concerns raised by animal welfare groups and the community (Bergmann, 2020). Many researchers and equestrians alike find the language of learning theory confusing (McLean & Christensen, 2017); a study of racehorse trainers from the UK reported trainers were put off by scientific language (Richardson et al., 2020). So, accessibility of scientific language may be a factor. Further, the same study reported that racehorse trainers were skeptical that a scientist could help their horse run faster. Others have reported that equestrians are put off by the “science” aspects of equitation science, which is underpinned by learning theory, reporting it is too rigid and mechanistic, portrays the horse as similar to a machine, and does not adequately account for the complexity of individual horse–human interactions (Haigh & Thompson, 2015). While equestrians’ perceptions of science and learning theory may or may not be accurate, the concerns they raise provide useful insights into how scientists might proceed to overcome these barriers and improve the relevance of science to equestrians. Increasing the relevance of science and overcoming equestrians’ skepticism are likely critical to increasing equestrians’ adoption of scientifically based training methods and the level of mastery with which these methods are applied.

Examining the second question of why rider knowledge did not translate into welfare or safety benefits, it could be that the welfare assessment was not sensitive enough to reveal a difference; however, if this the case, it may mean the difference is so small as to be rendered meaningless. Alternatively, it could be due to poor application of learning theory and/or the training goal being inconsistent with horse welfare. Correct application of learning theory is likely critical to it delivering the expected benefits of improved horse welfare and rider safety. Skepticism among equestrians about the efficacy of learning theory may lead them to invest minimal time into acquiring the knowledge and skills needed to apply learning theory successfully. Insufficient understanding and poor skill lead to poor results, thereby creating a self-fulfilling prophecy. Alternatively, a rider may have a thorough understanding of learning theory and excellent application skills, but the training goal is inconsistent with horse welfare. For example, the practice of hyperflexing the horse’s neck is widely used and popular in equestrian sport, yet there is evidence to suggest this practice compromises horse welfare (Borstel et al., 2009; Christensen et al., 2014; Mellor & Beausoleil, 2017). It is possible that a rider with excellent knowledge of learning theory could skillfully achieve this posture using light and sophisticated aids based on learning theory. However, because the training goal itself (hyperflexion) is inconsistent with good horse welfare, the mechanism by which the goal is achieved becomes irrelevant because poor horse welfare is always the outcome. Learning theory arises from a traditional reductionist scientific paradigm. Such a paradigm tends to reduce the horse to an object and encourages a narrow focus, two factors that place horses’ welfare at risk (Luke, Rawluk, et al., 2022). Poor application of learning theory due to perceived lack of relevance and skepticism of equestrians and/or setting training goals that are inconsistent with horse welfare are possible explanations for rider knowledge of learning theory not delivering expected horse welfare benefits.

The task of improving horse welfare and rider safety in equestrian sport is challenging owing to the size and complexity of the problem. Traditionally, scientists have taken a knowledge deficit approach to rectify equestrians’ lack of understanding of essential concepts, such as learning theory (Warren-Smith & McGreevy, 2008). However, it is recognized by many that while addressing knowledge deficits are necessary to improve horse welfare, it is unlikely they are sufficient, and strategies such as human behavior change will likely play an important role if significant improvements are to be achieved (Carroll et al., 2021). A key factor in driving human behavior change is engaging with the intended audience, in this case equestrians. Evidence from the education literature suggests engagement is enhanced when a student perceives a lesson has personal value and is relevant (Jang, 2008). If equestrians perceive scientific approaches have little value and relevance to horse keeping and training, education that does not address these issues is unlikely to succeed in creating change. Developing a new training framework that addresses equestrians’ concerns and encourages training to be seen in a larger context that is consistent with modern conceptualizations of animal welfare (Mellor et al., 2020) might offer a new pathway to increase equestrians’ adoption of scientifically based approaches.

To address equestrians’ concerns, a new horse-training framework would need to see the horse as a being rather than as an object, offer the flexibility of a wide-angle view of the horse to protect horse welfare, combined with the ability to zoom in and use a narrow focus to achieve specific training goals. To address equestrians’ objections, it must also offer nuanced solutions that account for the complexity of individual horse–human interactions. One such approach is a horse-training framework based on systems thinking (Capra & Luisi, 2014; Luke, Rawluk, et al., 2022). Such a framework would invite interdisciplinarity and partnership between scientists, riders, and horses, possibly making science feel more inclusive and more relevant to equestrians. Systems thinking helps explicate and articulate connections and subtleties that might otherwise be hidden. In doing so, a systems-thinking framework might allow equestrians to see more clearly the welfare issues faced by ridden horses and in turn, recognize the need for change. Offering equestrians a framework that is fundamentally holistic and inclusive, but within which traditional mechanistic science can be embedded, might address the concerns of equestrians that science does not “get” them and their horse, thus creating a pathway for scientists to make science more “approachable” and increase equestrians’ uptake of scientific approaches.

Community tolerance for poor sport-horse welfare is quickly disappearing (Ingle, 2021; Taylor, 2022). So, if equestrian sport is to have a long-term future, the need for strategies to deliver better horse welfare is high. A framework that overcomes equestrians’ barriers to science, then embedding learning theory into it, might see more widespread and faster adoption of training practices based on learning theory, along with more skillful application. Clearly, the development and testing of such a horse-training framework are needed. While systems thinking is yet to become established in the animal welfare literature, it has been identified as one the five key academic competencies needed to address perhaps the world’s greatest challenge, global sustainability (Wiek et al., 2011). As such, a horse-training framework based on systems thinking may stimulate new ideas and new directions for solving the problem of poor sport-horse welfare. The intention of this proposed framework is to facilitate discussion and help researchers find workable solutions to address the dual issues of poor horse welfare and poor human safety.

Acknowledgements

The authors acknowledge the generosity of the Australian horse riders who volunteered and completed our detailed survey. The authors also thank the anonymous peer reviewers for their thoughtful feedback on an earlier version of our manuscript.

Disclosure Statement

No potential conflict of interest was reported by the authors.