Behavioural signs of pain in the horse

The recognition of behavioural signs of pain in the horse is key to answering the first of the questions posed by Dawkins (Reference Dawkins2004): is the animal healthy? The importance of identifying signs that the horse is experiencing or anticipating pain was reflected in the number of articles on this topic that resulted from the search terms. A total of twenty-two articles relating to behavioural signs of pain associated with either specific health issues or in general were retained. Pain was induced in five of the studies, with the aim of identifying behavioural signs to facilitate the recognition of pain for future application. Behaviour recorded in four of these studies (Bussières et al. Reference Bussières, Jacques, Lainay, Beauchamp, Leblond, Cadoré, Desmaizières, Cuvelliez and Troncy2008; Grint et al. Reference Grint, Beths, Yvorchuk-St Jean, Whay and Murrell2017; Reid et al. Reference Reid, Rogers, Gronqvist, Gee and Bolwell2017; Egan et al. Reference Egan, Kearney, Brama, Parnell and McGrath2021) is shown in Table 2. In the study by Carvalho et al. (Reference Carvalho, Trindade, Conde, Antonioli, Funnicelli, Dias, Canola, Chinelatto and Ferraz2022) induced, low-grade, inflammatory pain was not detectable via analysis of facial expression (and consequently is not included in Table 2). Low-level pain was not found to affect typical feeding behaviour, but behavioural variability increased as horses recovered from induced bilateral lameness resulting from mild inflammation in the joints, although this varied for individual horses (Egan et al. Reference Egan, Kearney, Brama, Parnell and McGrath2021). There was also variation in individual behavioural responses to the induced pain and recovery, with no generic ‘normal’ behaviour identified (Egan et al. Reference Egan, Kearney, Brama, Parnell and McGrath2021).

Table 2. Behavioural signs of pain in the horse (in relation to specific health issues and in general)

The differentiation between behaviour resulting from anxiety or pain was explored by Reid et al. (Reference Reid, Rogers, Gronqvist, Gee and Bolwell2017), by testing responses to anxiety-provoking situations (social isolation) and/or a pain-eliciting stimulus (neck pinch). Pain alone resulted in reduced locomotion and reduced contact-seeking behaviour, whereas anxiety resulted in increased locomotion, increased vocalisation, restlessness, and contact-seeking behaviour. Anxiety and pain together caused restlessness. The authors concluded that social isolation (and potentially other sources of anxiety) may affect behaviour associated with pain responses, or potentially be the cause of behaviour typically cited as being indicative of pain (Reid et al. Reference Reid, Rogers, Gronqvist, Gee and Bolwell2017).

Behavioural signs of pain were also found to be confounded by associated stressors by Erber et al. (Reference Erber, Wulf, Becker-Birck, Kaps, Aurich, Möstl and Aurich2012). A comparison between the effect of hot branding and microchipping (as a means of identification) on the behavioural and physiological responses of foals concluded that both processes were stressful and that the associated restraint was a major stressor. No specific behavioural details were provided but the authors noted that painful procedures should always be avoided, particularly in young animals where early experiences may have an impact both on future pain perception, and on responses to handling and other management procedures (Erber et al. Reference Erber, Wulf, Becker-Birck, Kaps, Aurich, Möstl and Aurich2012). Dalla Costa et al. (Reference Dalla Costa, Pascuzzo, Leach, Dai, Lebelt, Vantini and Minero2018) found that certain facial expressions included in the Horse Grimace Scale (developed as a means of identifying pain in horses) were also associated with fear-eliciting situations (stiffly backward ears and prominent strained chewing muscles). Given that facial characteristics vary according to breed and type of horse (Hintze et al. Reference Hintze, Smith, Patt, Bachmann and Würbel2016) and Carvalho et al. (Reference Carvalho, Trindade, Conde, Antonioli, Funnicelli, Dias, Canola, Chinelatto and Ferraz2022) found that low-grade pain was not identifiable from facial expression, interpretation of these subtle behavioural signs should be treated with caution.

Evidence suggests that anticipation of pain and pain in general results in immobility and the avoidance of human contact (Pritchett et al. Reference Pritchett, Ulibarri, Roberts, Schneider and Sellon2003; Burn et al. Reference Burn, Dennison and Whay2010; Fureix et al. Reference Fureix, Menguy and Hausberger2010; Reid et al. Reference Reid, Rogers, Gronqvist, Gee and Bolwell2017). Severe and/or acute pain was recognised by the horses’ behavioural attempts at avoiding contact with humans (Burn et al. Reference Burn, Dennison and Whay2010; Fureix et al. Reference Fureix, Menguy and Hausberger2010), lack of any movement likely to worsen the pain (Pritchett et al. Reference Pritchett, Ulibarri, Roberts, Schneider and Sellon2003; Reid et al. Reference Reid, Rogers, Gronqvist, Gee and Bolwell2017) and decreased time spent eating (Trindade et al. Reference Trindade, Taffarel and Luna2021). Monitoring any changes in these features of behaviour can provide important insights into the ongoing health status of the horse.

Certain behaviours have been shown to be associated with specific health issues (as shown in Table 2), but individual variation occurs. For example, although Oliveira et al. (Reference Oliveira, Santos, Silva, Trindade, Yamada, Jaramillo, Silva and Baccarin2022) found that the severity of osteoarthritis affected the time horses spent in recumbency during hospitalisation (see Table 2), Kelemen et al. (Reference Kelemen, Grimm, Long, Auer and Jenner2021) noted that neither age nor lameness due to chronic orthopaedic disease had an impact on the time horses spent recumbent.

Behavioural signs of pain in the horse (in relation to specific health issues and in general) are shown in Table 2. See Table S2.1 in the Supplementary material for details of each study.

Tests of emotional reactivity in the horse

Challenges to human safety during interactions with horses, both during handling and ridden work, have been associated with sudden/unexpected, often intense, behavioural responses of the horse (Górecka-Bruzda et al. Reference Górecka-Bruzda, Jastrzębska, Gajewska, Muszyńska and Pieniąże2015; Lansade et al. Reference Lansade, Bonneau, Parias and Biau2019). Tests of emotional reactivity to various acute stressors (for example, social separation/isolation, startling/sudden occurrences, and novelty) have been developed to explore factors affecting such behavioural responses, with the aim of enabling better assessment and selection of horses and reducing human risk. Nineteen articles reporting the results of experimental tests of emotional reactivity, including novel object, novel environment, startle, and isolation tests, were retained. Factors found to affect reactivity were age (Baragli et al. Reference Baragli, Banti, Vitale and Sighieri2014; Bulens et al. Reference Bulens, Sterken, Van Beirendonck, Van Thielen and Driessen2015; Lee et al. Reference Lee, Kim, Lee and Kim2021), breed/type (Lee et al. Reference Lee, Kim, Lee and Kim2021), type of work/role (Hausberger et al. Reference Hausberger, Muller and Lunel2011; Mendonça et al. Reference Mendonça, Bienboire-Frosini, Kowalczyk, Leclercq, Arroub and Pageat2019), the effect of company (Lansade et al. Reference Lansade, Neveux and Levy2012; Rørvang & Christensen Reference Rørvang and Christensen2018; Ricci-Bonot et al. Reference Ricci-Bonot, Romero, Nicol and Mills2021), training (Squibb et al. Reference Squibb, Griffin, Favier and Ijichi2018) and laterality (Larose et al. Reference Larose, Richard-Yris, Hausberger and Rogers2006; De Boyer Des Roches et al. Reference De Boyer Des Roches, Richard-Yris, Henry, Ezzaouïa and Hausberger2008; Baragli et al. Reference Baragli, Scopa, Felici and Reddon2021). As above, individual differences in the behavioural reactions elicited by these tests were reported by Lansade et al. (Reference Lansade, Bouissou and Erhard2008), Pérez Manrique et al. (Reference Pérez Manrique, Hudson, Bánszegi and Szenczi2019) (see also Behaviour associated with weaning in Part 1; Hall & Kay Reference Hall and Kay2024), Safryghin et al. (Reference Safryghin, Hebesberger and Wascher2019), and Manrique et al. (Reference Manrique, Bánszegi, Hudson and Szenczi2021).

Behaviour of the horse during handling and management procedures

There were seventeen articles retained that had a focus on handling and management procedures. The predominant behavioural measures used were the extent to which the horse tried to avoid the procedure or was relaxed/approached the human carrying out the procedure, and/or whether the horse showed behavioural signs indicative of stress. Approach/avoidance behaviours were concluded as signifying positive or negative responses, respectively. Behavioural signs of stress were identified based on past studies (for example, Young et al. Reference Young, Creighton, Smith and Hosie2012) and were believed to indicate negative affective state. Behaviour indicative of positive or negative responses to handling and management procedures are shown in Table 3(a). Included in Table 3 are behaviours during handling tests that were associated with different home environments (Harewood & McGowan Reference Harewood and McGowan2005; Yarnell et al. Reference Yarnell, Hall, Royle and Walker2015) See Part I of this review, The home environment (Hall & Kay Reference Hall and Kay2024).

Table 3. (a) Behaviour during handling and management procedures indicative of affective state (positive or negative), supporting evidence*, and b) factors affecting this behaviour

* Key to supporting evidence: Past studies (PS), physiological measures (PHYS), assumption of associated pain (PIP), situations deemed positive or negative (+ve/-ve SIT), positive or negative social interactions (horse/human) (+ve/-ve SOC), approach/avoidance (APP/AVO), choice (PREF).

Behaviour of the horse when ridden

A total of seventeen articles with a focus on ridden behaviour were retained. These included studies assessing the effect of head/neck position on behaviour when ridden (von Borstel et al. Reference von Borstel, Duncan, Shoveller, Merkies, Keeling and Millman2009; Christensen et al. Reference Christensen, Beekmans, van Dalum and Van Dierendonck2014), comparative approaches to training (Visser et al. Reference Visser, Van Dierendonck, Ellis, Rijksen and Van Reenen2009; Mendonça et al. Reference Mendonça, Bienboire-Frosini, Sanchez, Kowalczyk, Teruel, Descout and Pageat2020), potential stress associated with therapeutic ridden sessions (Kaiser et al. Reference Kaiser, Heleski, Siegford and Smith2006; McDuffee et al. Reference McDuffee, Carr and Montelpare2022), the effect of different riders (Christensen et al. Reference Christensen, Munk, Hawson, Palme, Larsen, Egenvall and Rørvang2021) and behavioural indicators of pain (Dyson & Pollard Reference Dyson and Pollard2021) or anticipated pain (Dyson et al. Reference Dyson, Bondi, Routh, Pollard, Preston, McConnell and Kydd2022) associated with ridden work (see also Behavioural signs of pain in the horse). In most studies, the behaviour being assessed was that which relates to how the horse responds to the rider’s signals. The focus was on behaviour that deviated from that desired by the rider, being referred to as indicative of stress, discomfort, fear, frustration, and conflict (Kaiser et al. Reference Kaiser, Heleski, Siegford and Smith2006; Visser et al. Reference Visser, Van Dierendonck, Ellis, Rijksen and Van Reenen2009; von Borstel et al. Reference von Borstel, Duncan, Shoveller, Merkies, Keeling and Millman2009; Christensen et al. Reference Christensen, Beekmans, van Dalum and Van Dierendonck2014; Jastrzębska et al. Reference Jastrzębska, Wolska, Minero, Ogłuszka, Earley, Wejer and Górecka-Bruzda2017; Ruet et al. Reference Ruet, Biau, Arnould, Galloux, Destrez, Pycik, Boichot and Lansade2020; Christensen et al. Reference Christensen, Munk, Hawson, Palme, Larsen, Egenvall and Rørvang2021; McDuffee et al. Reference McDuffee, Carr and Montelpare2022). Consequently, the behaviours indicative of negative affect when the horse is ridden as shown in Table 4(a) far outweigh behavioural signs of positive affect (which, if the same criteria were applied as for negative behavioural signs, would presumably be calm acceptance of rider signals and delivery of the requested behavioural response).

Table 4. (a) Behaviour during ridden activities indicative of affective state (positive or negative), supporting evidence*, and (b) factors affecting this behaviour

* Key to supporting evidence: Past studies (PS), physiological measures (PHYS), assumption of associated pain (PIP), situations deemed positive or negative (+ve/-ve SIT), positive or negative social interactions (horse/human) (+ve/-ve SOC), approach/avoidance (APP/AVO), choice (PREF).

The most common justification for assigning valence to the behaviours recorded was reference to previous studies and to commonly accepted interpretations of the responses to rider signals during different ridden scenarios. Of the sources cited as informing the ethograms used, only five were cited by more than one study. McDonnell (Reference McDonnell2003) and Waring (Reference Waring2003) were each cited by three separate studies, and Young et al. (Reference Young, Creighton, Smith and Hosie2012), McGreevy et al. (Reference McGreevy, McLean, Warren-Smith, Waran and Goodwin2005), and Dyson et al. (Reference Dyson, Berger, Ellis and Mullard2018) were each cited by two separate studies (although the latter was only cited to support further studies by the same research group).

Ruet et al. (Reference Ruet, Biau, Arnould, Galloux, Destrez, Pycik, Boichot and Lansade2020) compared ridden performance with behavioural indicators of welfare state in the home environment, using both objective behavioural assessment and the more subjective qualitative behaviour assessment (QBA). They found an association between horses recorded as hypervigilant and/or performing stereotypical behaviours in the stable, and both signs they attributed to negative affective state when ridden (ear and tail positions) and QBA scores (Ruet et al. Reference Ruet, Biau, Arnould, Galloux, Destrez, Pycik, Boichot and Lansade2020). Out of the eleven behavioural and postural indicators used to assess affective state when ridden, only three could potentially be considered as positive signs (ears forwards or asymmetric, and snorts). Within the QBA list of thirteen descriptors, six could be described as positive in some scenarios (at ease, curious, friendly, happy, looking for contact, relaxed), although may not be appropriate for assessing ridden behaviour. No association between behaviour in the ridden test and QBA scores was reported (Ruet et al. Reference Ruet, Biau, Arnould, Galloux, Destrez, Pycik, Boichot and Lansade2020).

The challenges associated with attributing affective state to specific behaviours were highlighted by von Borstel et al. (Reference von Borstel, Duncan, Shoveller, Merkies, Keeling and Millman2009) where the inclusion of the vocalisation termed a ‘snort’ as a negative sign was supported by Waring (Reference Waring2003) but deemed a positive sign by Fraser (Reference Fraser1998). Similarly, Visser et al. (Reference Visser, Van Dierendonck, Ellis, Rijksen and Van Reenen2009) interpreted snorts as negative signs in contrast to Stomp et al. (Reference Stomp, Masson, Henry, Hausberger and Lesimple2020). It is possible that subtle differences in acoustic structure relate to valence, but these may be hard (or impossible) for the human ear to distinguish. More conclusive evidence of the subjective experience of the horse when ridden was provided by von Borstel et al. (Reference von Borstel, Duncan, Shoveller, Merkies, Keeling and Millman2009) in the form of a choice test. Horses showed a preference for an area where they had not been ridden in a coercively obtained head/neck position, with the latter hyperflexion also being found to be associated with increased fear-related responses (von Borstel et al. Reference von Borstel, Duncan, Shoveller, Merkies, Keeling and Millman2009).

Physiological measures were recorded in ten of the retained studies. Christensen et al. (Reference Christensen, Munk, Hawson, Palme, Larsen, Egenvall and Rørvang2021) found increases in HR and salivary cortisol concentration associated with specific riders. Visser et al. (Reference Visser, Van Dierendonck, Ellis, Rijksen and Van Reenen2009) recorded higher HR in horses that had been trained using conventional as opposed to more sympathetic methods in the final ridden test, although there was no difference in behaviour or performance. When comparing competitive success in showjumping with levels of salivary cortisol and occurrences of conflict behaviour, increased levels of both were associated with poorer performance (Jastrzębska et al. Reference Jastrzębska, Wolska, Minero, Ogłuszka, Earley, Wejer and Górecka-Bruzda2017). Whether such differences relate to physical exertion and/or fitness, or whether they reflect ‘mental stress’, is questionable. It has been noted previously that although physiological measures provide information about arousal levels, the extent to which they can be used to determine the valence of responses has yet to be determined, particularly in scenarios involving physical exertion (Hall et al. Reference Hall, Randle, Pearson, Preshaw and Waran2018).

It has been demonstrated that it is possible to adopt a responsive approach to the management procedure of grooming, which makes it a positive experience for the horse (Lansade et al. Reference Lansade, Bonneau, Parias and Biau2019). A similar more responsive approach to other management procedures, making them less aversive and pleasurable where possible, would reduce the cumulative effect of negative experiences for the horse. The same applies to ridden work where the rider was found to be the main factor associated with signs of conflict in the horse (Christensen et al. Reference Christensen, Munk, Hawson, Palme, Larsen, Egenvall and Rørvang2021). Although the frequency of conflict behaviours has been associated with poor performance in showjumping (Jastrzębska et al. 2017), the occurrence of conflict behaviour was not found to relate to performance evaluation by dressage judges (Hamilton et al. Reference Hamilton, Lancaster and Hall2022). In addition to recommending further training for riders, adapting performance criteria to account for behavioural signs of negative affective state would be positive progress towards improved welfare for ridden horses.

The potential for using behavioural signs to assess affective state in this scenario is limited by the restraint/control exerted by the rider and/or equipment used in ridden work. Also, when horses are kept in constrained home environments (stables) any apparent enthusiasm for exercise when being ridden may be attributed to rebound behaviour. Consequently, behavioural signs of positive or negative anticipation during the preparation for ridden work may provide a more accurate means of assessment. Avoidance behaviours during tacking up and mounting, suggestive of negative anticipation, were found to be common in both leisure and sports horses by Dyson et al. (Reference Dyson, Bondi, Routh, Pollard, Preston, McConnell and Kydd2022). Underlying pain or discomfort during ridden work must be eliminated as the potential cause of this behaviour before considering the need for re-training. Restrictive equipment, often used to mask ‘unwanted’ behaviours (for example, tight nosebands used to prevent mouth movement), undoubtedly causes discomfort, with rebound behaviours occurring once the restriction is removed (Fenner et al. Reference Fenner, Yoon, White, Starling and McGreevy2016). Attempts to avoid the oral pressure of a bit (Eisersiö et al. Reference Eisersiö, Yngvesson, Hartmann and Egenvall2023) and behaviour recorded when the horse’s body is forced and fixed into unnatural positions (Smiet et al. Reference Smiet, Van Dierendonck, Sleutjens, Menheere, van Breda, de Boer, Back, Wijnberg and Van Der Kolk2014) are further examples of how some aspects of ridden work and training can be very negative experiences for the horse. The impact of ridden work on the affective state of the horse can be assessed not only by ridden responses, but also by anticipatory and post-ridden behaviour. Improved welfare in ridden horses can only be achieved if their behavioural signs of confusion, discomfort or pain are responded to, not by restrictive equipment or harsher training methods, but by identifying the underlying cause.

Factors found to affect behaviour of the horse when ridden are shown in Table 4(b). Study details for the articles relating to this scenario are provided in Table S2.4 in the Supplementary material.

Behaviour of the horse during non-procedural horse-human interactions

Horse-human interactions not related to specific management procedures or training were considered separately. The retained articles comprised thirteen in which factors that influence how horses respond to humans in general were investigated. Methods used to assess horse-human interactions included human approach tests, behavioural responses to visual, auditory, and olfactory signs of human emotion, and the effect of watching what were considered by the authors to be positive (grooming) or negative (veterinary intervention) horse-human interactions on behaviour. In the human approach tests the response to the human was assessed according to whether the horse behaved in a way that increased or decreased the distance between horse and human. Tests used included those that allowed the horse to freely choose whether to approach a stationary human (also termed a voluntary human approach test) (Sankey et al. Reference Sankey, Richard-Yris, Henry, Fureix, Nassur and Hausberger2010; Brubaker et al. Reference Brubaker, Schroeder, Sherwood, Stroud and Udell2021; Lerch et al. Reference Lerch, Cirulli, Rochais, Lesimple, Guilbaud, Contalbrigo, Borgi, Grandgeorge and Hausberger2021) or those where the human approached the horse (forced human approach test) (Minero et al. Reference Minero, Dalla Costa, Dai, Canali, Barbieri, Zanella, Pascuzzo and Wemelsfelder2018). Measures used to assess the response of horses (movement of horse in relation to the human and behavioural interactions with the human) in these tests are shown in Table 5(a)(i).

Table 5. (a) (i, ii, iii) Behaviour of the horse during non-procedural horse-human interactions indicative of affective state (positive or negative), (i) during behavioural tests, (ii) in response to human expressions of emotion, and (iii) to watching third party horse-human interactions, supporting evidence* (see footnote below table for an explanation of the abbreviations used), and (b) factors affecting this behaviour

* Key to supporting evidence: Past studies (PS), physiological measures (PHYS), assumption of associated pain (PIP), situations deemed positive or negative (+ve/-ve SIT), positive or negative social interactions (horse/human) (+ve/-ve SOC), approach/avoidance (APP/AVO), choice (PREF).

The ability of horses to recognise signs of human emotion (visual, auditory, and olfactory) was demonstrated in eight studies. Responses to human facial expressions depicting positive and negative human emotions (including happy, angry, disgust, and neutral expressions) were recorded (from photographs: Smith et al. Reference Smith, Proops, Grounds, Wathan and McComb2016; live: Baba et al. Reference Baba, Kawai and Takimoto-Inose2019; Merkies et al. Reference Merkies, Sudarenko and Hodder2022). The impact of facial expression, as well as non-verbal auditory stimuli (laughter, growling: Smith et al. Reference Smith, Proops, Grounds, Wathan, Scott and McComb2018a), human body posture (Smith et al. Reference Smith, Wilson, McComb and Proops2018b) and human body odour (fear, non-fear: Sabiniewicz et al. Reference Sabiniewicz, Tarnowska, Świątek, Sorokowski and Laska2020) on horse behaviour is shown in Table 5(a)(ii). Approach behaviour was more likely to occur in response to positive human facial expressions (Smith et al. Reference Smith, Proops, Grounds, Wathan and McComb2016; Baba et al. Reference Baba, Kawai and Takimoto-Inose2019; Merkies et al. Reference Merkies, Sudarenko and Hodder2022), positive non-verbal communication (laughter: Smith et al. Reference Smith, Proops, Grounds, Wathan, Scott and McComb2018a), and when the human adopted a submissive rather than a dominant body posture (Smith et al. Reference Smith, Wilson, McComb and Proops2018b). Human body odour associated with fear resulted in increased head raising and touching the human (Sabniewicz et al. 2020). Smith et al. (Reference Smith, Proops, Grounds, Wathan and McComb2016) recorded increased HR in response to the photographs of angry (as opposed to happy) human faces, whereas Merkies et al. (Reference Merkies, Sudarenko and Hodder2022) found when using live actors (with facial expressions of anger, sadness, joy, and neutral) there was no effect on the horse’s HR.

Further evidence of the horse’s ability to recognise the valence associated with human expression was demonstrated in two studies investigating behavioural responses to incongruous cross-modal signals (facial expression and vocalisation: Nakamura et al. Reference Nakamura, Takimoto-Inose and Hasegawa2018; Trösch et al. Reference Trösch, Cuzol, Parias, Calandreau, Nowak and Lansade2019). In both studies, the horses reacted in accordance with the valence of the vocalisation, particularly if this was negative and incongruent with the visual image. Nakamura et al. (Reference Nakamura, Takimoto-Inose and Hasegawa2018) recorded an increase in HR and length of time spent looking at the visual image if the negative vocalisation (scolding voice) was from a familiar human and incongruent with the facial expression on the image. Trösch et al. (Reference Trösch, Cuzol, Parias, Calandreau, Nowak and Lansade2019) also recorded increased HR and attention paid to an incongruent image depicting facial expression when the valence of the human vocalisation was negative (phonetic sound representing anger). See also Part I of this review (Hall & Kay Reference Hall and Kay2024; Intra-species social behaviour) where the salience of vocalisation as a means of communicating emotion between horses was discussed.

The potential for emotional contagion to occur when horses watched horse-human interactions on a screen was demonstrated by Trösch et al. (Reference Trösch, Pellon, Cuzol, Parias, Nowak, Calandreau and Lansade2020). Responses to watching interactions deemed positive (grooming) differed from those recorded when watching negative (veterinary procedures) horse-handler interactions, with contact-seeking behaviour and lower HR associated with the positive interactions. However, following the viewing, when presented with both handlers, increased contact-seeking behaviour was associated with the negatively interacting handler which the authors tentatively suggested could be linked to attempts at appeasement, but the reason was not clear.

What was evident from the findings of the retained articles was the ability of the horse to respond to human expressions of emotion, with preference generally given for positive signals. It is recognised that human emotion is contagious (Herrando & Constantinides Reference Herrando and Constantinides2021), and the attempted mutual grooming behaviour elicited by watching video footage of positive horse-human interactions during grooming (Trosch et al. 2020), suggests that emotional transfer can occur between human and horse (see Table 5[a [iii]).

Factors found to affect horse behaviour during non-procedural horse-human interactions are shown in Table 5(b).

Study details for the articles relating to this scenario are provided in Table S2.5 in the Supplementary material.

Pre-transport

It was reported that the loading phase of the journey was found to be the most stressful (Siniscalchi et al. Reference Siniscalchi, Padalino, Lusito and Quaranta2014), with evasive and avoidance behaviour being used as a measure of this (Waran & Cuddeford Reference Waran and Cuddeford1995; Hendriksen et al. Reference Hendriksen, Elmgreen and Ladewig2011; Siniscalchi et al. Reference Siniscalchi, Padalino, Lusito and Quaranta2014). Increases in HR were recorded during loading into both road (Waran & Cuddeford Reference Waran and Cuddeford1995; Tateo et al. Reference Tateo, Padalino, Boccaccio, Maggiolino and Centoducati2012; Siniscalchi et al. Reference Siniscalchi, Padalino, Lusito and Quaranta2014) and air transport vehicles (Stewart et al. Reference Stewart, Foster and Waas2003; Munsters et al. Reference Munsters, De Gooijer, Van Den Broek and Van Oldruitenborgh-Oosterbaan2013). Although Hendriksen et al. (Reference Hendriksen, Elmgreen and Ladewig2011) found that training to load horses using positive reinforcement resulted in a reduction in behavioural signs of discomfort (compared with using negative reinforcement), this was not associated with any difference in HR. Although the increases in HR may, in part, reflect increased arousal associated with anticipation (positive or negative), it is likely that the primary cause of this was the physical exertion involved in entering the vehicle.

During transport

When the transport vehicle was moving horses spent more time in a braced position and leaning against the partition, and less time feeding than when the vehicle was stationary (Waran & Cuddeford Reference Waran and Cuddeford1995). Vehicle movement was associated with an increase in HR (Waran & Cuddeford Reference Waran and Cuddeford1995; Padalino & Raidal Reference Padalino and Raidal2020), which was also recorded during air transport at times of take-off and landing, but not throughout the rest of the flight (Stewart et al. Reference Stewart, Foster and Waas2003; Munsters et al. Reference Munsters, De Gooijer, Van Den Broek and Van Oldruitenborgh-Oosterbaan2013). Increases in HR were found to be reduced when the horse was transported with a live companion (as opposed to alone) (Kay & Hall Reference Kay and Hall2009), or when the horse travelled with its head in a lowered position (Padalino et al. Reference Padalino, Raidal, Knight, Celi, Jeffcott and Muscatello2018). Although the latter finding may relate to improved balance when in this position, isolation stress during transport was shown to be a contributory factor (Kay & Hall Reference Kay and Hall2009). In addition to a decrease in HR, travelling with a companion was also associated with a decrease in the behavioural signs of stress, including less time spent head tossing and vocalising, and more time spent eating than when travelling alone (Kay & Hall Reference Kay and Hall2009).

Post transport

Post-transport behaviour was recorded by Padalino et al. (Reference Padalino, Maggiolino, Boccaccio and Tateo2012), Tateo et al. (Reference Tateo, Padalino, Boccaccio, Maggiolino and Centoducati2012) and Ji et al. (Reference Ji, Yang, Ge, Wang, Cao and Hu2013). Following transportation for the first time (for 5 h, 15 mins), Przewalski horses (Equus ferus prezwalskii) moved more, stood resting less and drank less than pre-transport, and increases in faecal cortisol metabolites were recorded 24 h post-transport (returning to pre-transport levels within three days) (Ji et al. Reference Ji, Yang, Ge, Wang, Cao and Hu2013). After three-hour road journeys an increase in feeding and drinking was recorded by Padalino et al. (Reference Padalino, Maggiolino, Boccaccio and Tateo2012), and by Tateo et al. (Reference Tateo, Padalino, Boccaccio, Maggiolino and Centoducati2012). The confinement and frequently unfamiliar environment associated with transport has an impact on behaviour both during and after a journey. Most stress-related behaviours were recorded during the first hour of a journey (Padalino et al. Reference Padalino, Raidal, Knight, Celi, Jeffcott and Muscatello2018) and journeys of over three hours resulted in more rebound behaviour following transport confinement than shorter journeys (Tateo et al. Reference Tateo, Padalino, Boccaccio, Maggiolino and Centoducati2012).

See Table 6(b) for factors affecting horse behaviour during and post-transportation and Table S2.6 (Supplementary material) for details of the studies included in this scenario.

In summary

All stages of this activity have the potential to negatively affect the horse, over and above the physical challenges of remaining upright. Careful pre-planning, training, vehicle design and providing a travel companion, as well as allowing sufficient recovery time should all be travelling requirements. The fact that evasive and avoidance behaviour is so prevalent at the loading stage is evidence that many horses are at best apprehensive about entering confined vehicle spaces, particularly during initial training, and that measures that can be taken to ameliorate this should be applied.

Behaviour of the horse during training other than when ridden (methods and equipment)

Nine articles with a focus on training and groundwork, which did not include riding the horse, were retained. These included behavioural responses to foundation training (Nittynen et al. Reference Nittynen, Riihonen, Moscovice and Koski2022), traditional compared with natural horsemanship training (Fureix et al. Reference Fureix, Pagès, Bon, Lassalle, Kuntz and Gonzalez2009), responses associated with positively or negatively reinforced training (Sankey et al. Reference Sankey, Richard-Yris, Henry, Fureix, Nassur and Hausberger2010; Larssen & Roth Reference Larssen and Roth2022), different head and neck positions during lunging (Smiet et al. Reference Smiet, Van Dierendonck, Sleutjens, Menheere, van Breda, de Boer, Back, Wijnberg and Van Der Kolk2014), a low-intensity, stress-provoking task (walking backwards: Rietmann et al. Reference Rietmann, Stuart, Bernasconi, Stauffacher, Auer and Weishaupt2004b) and the effect of restrictive and/or pressure-inducing equipment (tight nosebands: Fenner et al. Reference Fenner, Yoon, White, Starling and McGreevy2016; rein tension/bit pressure: Eisersiö et al. Reference Eisersiö, Yngvesson, Byström, Baragli and Egenvall2021, Reference Eisersiö, Yngvesson, Hartmann and Egenvall2023). The measures used either focused on behavioural responses towards humans, reactivity in novel object and isolation tests, and signs of discomfort and/or rebound behaviour. Behaviours signifying positive and negative responses to training other than when ridden are shown in Table 7(a).

Table 7. (a) Behaviour of the horse during training other than when ridden indicative of affective state (positive or negative), supporting evidence* (see footnote below Table for an explanation of the abbreviations used), and (b) factors affecting this behaviour

* Key to supporting evidence: Past studies (PS), physiological measures (PHYS), assumption of associated pain (PIP), situations deemed positive or negative (+ve/-ve SIT), positive or negative social interactions (horse/human) (+ve/-ve SOC), approach/avoidance (APP/AVO), choice (PREF).

Following positive reinforcement training, increased approach/contact-seeking behaviour towards humans was recorded (Sankey et al. Reference Sankey, Richard-Yris, Henry, Fureix, Nassur and Hausberger2010; Larssen & Roth Reference Larssen and Roth2022), which may in part be the association of the human with food as this was used as the primary positive reinforcer during training. An association between an increase in HR and negative as opposed to positive reinforcement was found by Sankey et al. (Reference Sankey, Richard-Yris, Henry, Fureix, Nassur and Hausberger2010), but reinforcement type was not associated with differences in hair cortisol levels (Larssen & Roth Reference Larssen and Roth2022). However, the type of reinforcement used had no effect on overall affective state as determined by judgement bias testing (Larsson & Roth Reference Larssen and Roth2022).

Fureix et al. (Reference Fureix, Pagès, Bon, Lassalle, Kuntz and Gonzalez2009) found reduced latency to approach a motionless human in horses trained using natural horsemanship methods as opposed to traditional training, but although reduced responses to isolation and novel object tests occurred following training, this did not differ with training type (Fureix et al. Reference Fureix, Pagès, Bon, Lassalle, Kuntz and Gonzalez2009). As well as differences in response associated with the type/approach to training, a variation in how young horses respond to training was found by Nittynen et al. (Reference Nittynen, Riihonen, Moscovice and Koski2022). During foundation training, increases in salivary oxytocin occurred in horses showing affiliative, human-directed behaviours, with decreases in salivary oxytocin occurring in those showing signs of discomfort (Nittynen et al. Reference Nittynen, Riihonen, Moscovice and Koski2022). The authors suggest that salivary oxytocin may be a useful non-invasive indicator of the subjective experience of training on young horses (Nittynen et al. Reference Nittynen, Riihonen, Moscovice and Koski2022). In the same study, decreases in salivary cortisol concentration were recorded as the training progressed, alongside decreased signs of fear and discomfort, but the former increased with longer training sessions, suggestive of a link with physical exertion (Nittynen et al. Reference Nittynen, Riihonen, Moscovice and Koski2022). Horse-human interactions during training can undoubtedly have an impact on how horses perceive humans, but it is unclear the extent to which this contributes to more general affective state.

The exercise of backing up in hand was shown to be challenging for horses by Rietmann et al. (Reference Rietmann, Stuart, Bernasconi, Stauffacher, Auer and Weishaupt2004b), who found correlations between HR and HRV parameters and behaviour during this low intensity but mentally stressful exercise (head high, explosive deviation and stopping). Backing up was also used to investigate the relationship between rein tension and behaviour, and the effect of a bit during in-hand training (Eisersiö et al. Reference Eisersiö, Yngvesson, Byström, Baragli and Egenvall2021, Reference Eisersiö, Yngvesson, Hartmann and Egenvall2023). When a bit was used during the exercise (rather than a bit-less halter), an increase in head and neck movements, and oral behaviours was recorded (Eisersiö et al. Reference Eisersiö, Yngvesson, Byström, Baragli and Egenvall2021). Specific behaviours were found to affect rein tension, with increases of tension occurring with head movement down and forwards, decreased tension when the head was raised, the mouth open, and/or biting on the bit (Eisersiö et al. Reference Eisersiö, Yngvesson, Hartmann and Egenvall2023). Clear behavioural signs linked with attempts to relieve bit pressure on the mouth were recorded. It was also demonstrated that the correct application of negative reinforcement enabled rapid decreases in the rein pressure required to train the task (Eisersiö et al. Reference Eisersiö, Yngvesson, Byström, Baragli and Egenvall2021). Human training in the correct application of learning theory and an awareness of the behavioural signs of discomfort in the horse would both be beneficial in reducing the negative impact of equestrianism on horse welfare.

Restricting movement during training both masks the response of the horse and is likely to cause discomfort. Lunging horses with their head and necks maintained in a specific position (variations in elevation and flexion) resulted in attempted avoidance behaviour when the side reins were being attached and stepping backwards when fitted to maintain an elevated head position with the bridge of the nose around the vertical (Smiet et al. Reference Smiet, Van Dierendonck, Sleutjens, Menheere, van Breda, de Boer, Back, Wijnberg and Van Der Kolk2014). This latter position also resulted in the most conflict behaviour and increases in salivary cortisol concentration. Negative anticipatory behaviour was also recorded when side reins were fixed to maintain the position with the neck lowered and flexed, and the bridge of the nose was pointing towards the carpus (Smiet et al. Reference Smiet, Van Dierendonck, Sleutjens, Menheere, van Breda, de Boer, Back, Wijnberg and Van Der Kolk2014). The impact of noseband tightness on physiology and behaviour while standing in a test area for ten minutes was assessed by Fenner et al. (Reference Fenner, Yoon, White, Starling and McGreevy2016). In the tight noseband condition, an increase in HR, decreased HRV and an increase in eye temperature were recorded. Following the release of the pressure when the noseband was unfastened, yawning, swallowing, and licking (rebound behaviours) occurred (Fenner et al. Reference Fenner, Yoon, White, Starling and McGreevy2016). There is clear evidence that movement restriction is at best unnecessary during training, unhelpful in that it reduces the feedback that the horse can give to the human trainer, and at worst a negative and potentially painful experience for the horse.

Factors affecting behavioural responses during training other than when ridden and in response to the equipment used are shown in Table 7(b). See Table S2.7 in the Supplementary material.

General behavioural signs of affective state during interactions with humans

In the two-dimensional model of core affect proposed by Mendl et al. (Reference Mendl, Burman and Paul2010), positive valence when the animal is in a relatively high state of arousal is associated with behaviour motivated by reward acquisition. In a comparative state of arousal, negative valence is associated with behaviour motivated by the avoidance of punishment (and potentially pain/discomfort). There were specific examples of avoidance behaviour in the scenarios reported in this review. In ridden work, behaviour associated with restrictive practices (von Borstel et al. Reference von Borstel, Duncan, Shoveller, Merkies, Keeling and Millman2009; Christensen et al. Reference Christensen, Beekmans, van Dalum and Van Dierendonck2014) and behavioural signs of negative anticipation during preparation for being ridden (Dyson et al. Reference Dyson, Bondi, Routh, Pollard, Preston, McConnell and Kydd2022), again suggestive of avoidance attempts, should all be regarded as attempts by the horse to avoid negative consequences. During transport, avoidance behaviour was found to be prevalent at the loading stage (Waran & Cuddeford Reference Waran and Cuddeford1995; Hendriksen et al. Reference Hendriksen, Elmgreen and Ladewig2011; Siniscalchi et al. Reference Siniscalchi, Padalino, Lusito and Quaranta2014), potentially associated with the anticipation of entering an unknown environment and being separated from conspecifics. The occurrence of avoidance behaviour, in whatever context-specific form this takes, provides a clear signal that certain experiences are aversive to the horse. In contrast, examples of approach behaviour suggest that the horse anticipates a positive outcome. A tendency to approach a human displaying positive emotional signals as opposed to negative ones (for example, Smith et al. Reference Smith, Proops, Grounds, Wathan, Scott and McComb2018a,b; Merkies et al. Reference Merkies, Sudarenko and Hodder2022), suggests both the ability to recognise human expression and that this is used to predict a positive or negative outcome from subsequent interactions.

As well as providing context-specific information regarding the valence of emotional responses, approach/avoidance behaviour provides some insight into the horse’s overall affective state. Minero et al. (Reference Minero, Dalla Costa, Dai, Canali, Barbieri, Zanella, Pascuzzo and Wemelsfelder2018) found that horses assessed as being more at ease and relaxed in the home environment were most likely to readily approach a human when tested, and this behaviour in the home environment was judged to reflect a positive affective state. A positive association was found between the amount of forage provided and the tendency for horses to approach a human in a voluntary approach test (Lerch et al. Reference Lerch, Cirulli, Rochais, Lesimple, Guilbaud, Contalbrigo, Borgi, Grandgeorge and Hausberger2021).

In states of low arousal, Mendl et al. (Reference Mendl, Burman and Paul2010) describe the difference between positive and negative affective state as calm and relaxed or sad and depressed, respectively. It was noted in Part I of this review (Hall & Kay; The home environment) that differentiating between positive and negative signs when the horse is stationary may yield inconsistent conclusions, but general demeanour has the potential to provide an insight into the affective state of the horse. Hausberger et al. (Reference Hausberger, Fureix and Lesimple2016) acknowledged the need for robust, objective measures of physical health, and used postural features and stance to differentiate between standing resting, standing observing, and standing withdrawn. Behavioural variability, as observed to increase when horses were recovering from low-level pain (Egan et al. Reference Egan, Kearney, Brama, Parnell and McGrath2021), which includes periods of alertness and relaxation, would appear to be one important positive sign. With the proviso that subjective assessment is likely to be influenced by human experience, training, and role within the equine sector (as found in ridden horses by Hall et al. Reference Hall, Kay and Yarnell2014), a qualitative approach using evidence-based behavioural descriptors (similar to that used in the AWIN protocol: AWIN 2015), could be developed for wider application.

To enable an accurate response to the question posed by Dawkins (Reference Dawkins2004) regarding the health of the horse, behavioural signs of pain must be identified. Such signs should also determine the nature of subsequent interactions with humans, to ensure that these are not associated with increasing discomfort. Behavioural signs of negative anticipation may reflect underlying health issues or an association with past painful/negative experiences. Although it is not always easy to distinguish between behaviour indicative of pain and fear-related behaviour, and individuals vary in how both are expressed, this is an area that requires further investigation. Currently, pain-related behaviour may not always be interpreted as such, resulting in prolonged suffering, and potentially negatively affecting horse-human interactions.

Behaviour during many horse-human interactions is controlled/restricted by the human, and consequently signs indicative of positive or negative experiences may not be expressed by the horse and will also be affected by training. However, better recognition of behavioural signs of how horses experience interactions with humans would facilitate change in those identified as unpleasant for the horse.

The challenges and limitations of this systematic review (Parts I [Hall & Kay Reference Hall and Kay2024] and II combined)

The objectives of this systematic review, to identify evidence of how horse behaviour reflects affective state and its determining factors, with the overall aim of enabling the assessment of quality of life, were clear. However, assimilating the findings of the 179 disparate studies that were identified by the search terms and retained has proved to be complex. The variety of the scenarios explored reflects the diversity of activities and procedures that constitute the domestic life of the horse, at least in their role in sporting or leisure pursuits (as covered in this review). Notably, the findings suggest a complex inter-relationship between different aspects of the life of the domestic horse and the impact that they collectively have on their quality of life.

General challenges relating to research in horse welfare include accessibility/availability of subjects, consistency, and comparability of management conditions and ultimately the costs involved in studies involving the horse. Consequently, in many of the retained articles, low subject numbers are involved. A minimum for retention in this review was set at four subjects, although eight subjects per experimental group was suggested by Pongrácz (Reference Pongrácz2023) where statistical analyses were performed. The articles retained in this review included fourteen with fewer than eight subjects per experimental group (seven subjects: Rogers et al. Reference Rogers, Walsh, Gee and Firth2012; Smiet et al. Reference Smiet, Van Dierendonck, Sleutjens, Menheere, van Breda, de Boer, Back, Wijnberg and Van Der Kolk2014; Dalla Costa et al. Reference Dalla Costa, Bracci, Dai, Lebelt and Minero2017; Egan et al. Reference Egan, Kearney, Brama, Parnell and McGrath2021; six subjects: Harewood & McGowan Reference Harewood and McGowan2005; Werhahn et al. Reference Werhahn, Hessel and Van den Weghe2012; Ji et al. Reference Ji, Yang, Ge, Wang, Cao and Hu2013; Destrez et al. Reference Destrez, Grimm, Cézilly and Julliand2015; Reid et al. Reference Reid, Rogers, Gronqvist, Gee and Bolwell2017; Carvalho et al. Reference Carvalho, Trindade, Conde, Antonioli, Funnicelli, Dias, Canola, Chinelatto and Ferraz2022; five subjects: Marliani et al. Reference Marliani, Sprocatti, Schiavoni, Bellodi and Accorsi2021; Robinson & Bye Reference Robinson and Bye2021; and four subjects: Quick & Warren-Smith Reference Quick and Warren-Smith2009; McDuffee et al. Reference McDuffee, Carr and Montelpare2022). Most of the studies with low subject numbers involved repeated measures, an exception being Quick and Warren-Smith (Reference Quick and Warren-Smith2009). In only two articles, both included in Behavioural signs of pain in the horse, were references made to power calculations to establish the minimum required sample size (Dodds et al. Reference Dodds, Knight, Allen and Murrell2017; Ortolani et al. Reference Ortolani, Scilimati, Gialletti, Menchetti and Nannarone2021). Dodds et al. (Reference Dodds, Knight, Allen and Murrell2017) noted that a power calculation was not carried out prior to the start of the study because there was no preliminary data upon which to base such analyses, and data collection was dependent upon the number of eligible horses that presented to the clinic during the study period. Similar challenges were noted by Ortolani et al. (Reference Ortolani, Scilimati, Gialletti, Menchetti and Nannarone2021) who reported calculating the required sample size as 24 but ended up with only 23 subjects.

At least in part, it is the result of challenges relating to availability, controllability, and cost that this review is based on multiple, relatively small studies. It does however provide a broad overview of what is also a large and varied sector. However, one of the downfalls of the number of articles retained within this review has been the time that it has taken to evaluate and assimilate the results of 179 separate studies. The time lapse between the date of the search deadline and manuscript submission has undoubtedly resulted in the omission of the most recent articles, but those retained provide robust evidence of measures that must be taken to improve the quality of life of the domestic horse.

Within the retained articles only three examples of studies where preference tests had been included were identified (see Part I [Hall & Kay Reference Hall and Kay2024]; Table 2(b): preferred method of forage presentation in horse groups, Melvin et al. Reference Melvin, Costello and Colpoys2020; Part I (Hall & Kay Reference Hall and Kay2024); Table 3(b): preference for companion over food in mares, Górecka-Bruzda et al. Reference Górecka-Bruzda, Jastrzębska, Drewka, Nadolna, Becker and Lansade2022; and Part II; Behaviour of the horse when ridden: preference for a location associated with being ridden in a regular poll position as opposed to coercively obtained hyperflexion of the head and neck: von Borstel et al. Reference von Borstel, Duncan, Shoveller, Merkies, Keeling and Millman2009). Although constrained by human-selected options, further use of preference testing, and the comparison of behaviour in preferred and non-preferred situations could provide valuable insights into behavioural signs of positive affective state (with the proviso that individual differences in the behavioural expression of affective state will occur).

Animal welfare implications

Once the behavioural needs of the horse have been addressed, the impact of other aspects of management and training require consideration. As evidenced by the number of scenarios covered in this review, and factors that have been shown to have a negative impact on behaviour, it is likely that these will have a negative effect on the QOL of the domestic horse. As acknowledged in other species (for example, experimental cattle and pigs; Ryan et al. Reference Ryan, Waters and Wolfensohn2021), QOL is determined by the cumulative impact of past experiences, and the balance of pleasant and unpleasant experiences that horses have in their many and varied interactions with humans will contribute to their QOL. Whenever possible, management procedures should be designed to reduce aversion and adapted to be pleasurable for the horse, as demonstrated during grooming by Lansade et al. (Reference Lansade, Nowak, Lainé, Leterrier, Bonneau, Parias and Bertin2018). Also, there should be careful consideration of whether potentially aversive procedures are necessary and where possible their duration and/or frequency reduced or avoided altogether. Adherence to the LIMA (least invasive, minimally aversive) principles, as included in the Joint Standards of Practice and Code of Ethics adopted by the International Association of Animal Behaviour Consultants (IAABC) in 2018 (https://iaabc.org/code-of-ethics), should be used to guide all aspects of horse management and training. Clearly, some necessary procedures may be unavoidably aversive (for example, veterinary procedures) but any negative impact can be reduced by ensuring that the human handler is skilled in the procedure and that it is completed as quickly as possible (Górecka-Bruzda et al. Reference Górecka-Bruzda, Jaworski, Suwała, Sobczyńska, Jastrzębska, Ogłuszka, Sankey, Boroń and Jezierski2017). Familiarity with the handler (Marsbøll & Christensen Reference Marsbøll and Christensen2015; Liehrmann et al. Reference Liehrmann, Viitanen, Riihonen, Alander, Koski, Lummaa and Lansade2022), physical restraint (Guinnefollau et al. Reference Guinnefollau, Bolwell, Gee, Norman and Rogers2021), the environment in which the procedure is carried out and the level of training/habituation to the procedure, will all affect the experience of the horse. The potentially negative cumulative impact of repeated aversive procedures (as may be required in a veterinary context) must be acknowledged and measures taken to provide positive experiences to counterbalance this.

The extent to which these recommendations are applied will ultimately depend upon the decisions and behaviour of those responsible for caring for the horse. In agreement with the first principle of humane livestock farming compiled by The Council on Animal Affairs in The Netherlands, there must be a recognition of the intrinsic value of the animal as a sentient being that can experience pain and pleasure (Council on Animal Affairs 2021). Acknowledgement of the importance of providing the horse with sufficient opportunities to fulfil its natural behavioural needs, and recognition of the inextricable link between behavioural satisfaction and affective state, is the basis for moving closer to providing a good life for the domestic horse.

To promote positive affective state in the domestic horse there are key issues that need addressing, as shown in Figure 1. The first challenge is for those within the equine sector to acknowledge that change is needed. Changes for the better, however small, in any of the areas identified, have the potential to make a positive impact on the horse. Once horse behaviour is recognised more clearly as an indicator of affective state, and subsequently results in a change in human behaviour, the horse will have a better chance of living a good life. The outcome for the horse is reliant upon these changes in human behaviour.

Figure 1. Flow chart demonstrating how human behaviour can facilitate the progression from satisfying the behavioural needs of the horse to providing a good life for the domestic horse.