The home environment

The greatest number of articles retained in Part I of this review (21) focused on the behaviour of the horse in its home environment. The behaviour recorded varied according to the specific focus of the study and the features of the home environment that were under investigation. Features included stable type/size/design (Burla et al. Reference Burla, Rufener, Bachmann, Gygax, Patt and Hillmann2017; Lesimple et al. Reference Lesimple, Gautier, Benhajali, Rochais, Lunel, Bensaïd, Khalloufi, Henry and Hausberger2019; Ruet et al. Reference Ruet, Lemarchand, Parias, Mach, Moisan, Foury, Briant and Lansade2019; Marliani et al. Reference Marliani, Sprocatti, Schiavoni, Bellodi and Accorsi2021), paddock vs stabling/turn-out opportunities (Harewood & McGowan Reference Harewood and McGowan2005; Werhahn et al. Reference Werhahn, Hessel and Van den Weghe2012; Pessoa et al. Reference Pessoa, Trigo, Mesquita Neto, Lacreta Junior, Sousa, Muniz and Moura2016; Suagee-Bedore et al. Reference Suagee-Bedore, Linden and Bennett-Wimbush2021), the social environment (individual, paired or group housing; Harewood & McGowan Reference Harewood and McGowan2005; Visser et al. Reference Visser, Ellis and Van Reenen2008; Werhahn et al. Reference Werhahn, Hessel and Van den Weghe2012; Yarnell et al. Reference Yarnell, Hall, Royle and Walker2015), bedding area/depth (Burla et al. Reference Burla, Rufener, Bachmann, Gygax, Patt and Hillmann2017; Greening et al. Reference Greening, Downing, Amiouny, Lekang and McBride2021), mode of forage provision (Melvin et al. Reference Melvin, Costello and Colpoys2020; Sundman et al. Reference Sundman, Goodwin, Reisinger, Smith, Costello, Walter and Colpoys2022; Baumgartner et al. Reference Baumgartner, Erhard and Zeitler-Feicht2023) and environmental enrichment (Lansade et al. Reference Lansade, Valenchon, Foury, Neveux, Cole, Layé, Cardinaud, Lévy and Moisan2014, Reference Lansade, Lemarchand, Reigner, Arnould and Bertin2022).

The retained articles used one of two main approaches, either assessing the impact of current (long-term) housing or testing the effect of specific conditions set up experimentally for shorter periods. In five of the retained articles the effect of long-term housing type and the association between housing features and behavioural and physiological signs of welfare was assessed (Lesimple et al. Reference Lesimple, Gautier, Benhajali, Rochais, Lunel, Bensaïd, Khalloufi, Henry and Hausberger2019; Ribeiro et al. Reference Ribeiro, Matzkeit, Nicolau JTde, Castilha, de Oliveira and Bankuti2019; Ruet et al. Reference Ruet, Lemarchand, Parias, Mach, Moisan, Foury, Briant and Lansade2019; Sénèque et al. Reference Sénèque, Lesimple, Morisset and Hausberger2019; Sauveroche et al. Reference Sauveroche, Henriksson, Theodorsson, Svensson Holm and Roth2020). Where a timescale for the current housing was provided, this ranged from > 30 days (Ribeiro et al. Reference Ribeiro, Matzkeit, Nicolau JTde, Castilha, de Oliveira and Bankuti2019) to > 9 months (Ruet et al. Reference Ruet, Lemarchand, Parias, Mach, Moisan, Foury, Briant and Lansade2019). The impact of shorter-term exposure to differing housing situations was assessed either by comparing different horses kept under different conditions for a specific length of time (between subjects: 12 weeks; Visser et al. Reference Visser, Ellis and Van Reenen2008, five weeks; Lansade et al. Reference Lansade, Valenchon, Foury, Neveux, Cole, Layé, Cardinaud, Lévy and Moisan2014), or by repeated measures where all horses underwent all conditions each for a set time-period which varied from 1 h (Suagee-Bedore et al. Reference Suagee-Bedore, Linden and Bennett-Wimbush2021), 24 h (Harewood & McGowan Reference Harewood and McGowan2005), five days (Yarnell et al. Reference Yarnell, Hall, Royle and Walker2015), six days (Greening et al. Reference Greening, Downing, Amiouny, Lekang and McBride2021), eleven days (Burla et al. Reference Burla, Rufener, Bachmann, Gygax, Patt and Hillmann2017), 14 days (Werhahn et al. Reference Werhahn, Hessel and Van den Weghe2012; Melvin et al. Reference Melvin, Costello and Colpoys2020), 16 days (Pessoa et al. Reference Pessoa, Trigo, Mesquita Neto, Lacreta Junior, Sousa, Muniz and Moura2016), to approximately 34 days non-consecutively in experiment 2 of the study by Lesimple et al. (2019). Periods of adaptation to each condition were generally included, with behavioural and physiological assessments taking place towards the end of the period in each condition (for example, Burla et al. Reference Burla, Rufener, Bachmann, Gygax, Patt and Hillmann2017 recorded behaviour over the last three days of eleven-day treatment periods). Subject numbers included in the studies assessing long-term effects of housing ranged from 32 (experiment 1; Lesimple et al. Reference Lesimple, Gautier, Benhajali, Rochais, Lunel, Bensaïd, Khalloufi, Henry and Hausberger2019) to 187 where behaviour was recorded via scan sampling for 50 days over a nine-month period (Ruet et al. Reference Ruet, Lemarchand, Parias, Mach, Moisan, Foury, Briant and Lansade2019). Subject numbers included in the shorter-term studies ranged from six (Harewood & McGowan Reference Harewood and McGowan2005) to 38 (Burla et al. Reference Burla, Rufener, Bachmann, Gygax, Patt and Hillmann2017). See Table S1.1 in the Supplementary material for details relating to all studies retained in this scenario.

Behaviour recorded

Behaviour in the home environment indicative of affective state (positive or negative) and the type of supporting evidence reported is shown in Table 2(a). The impact of different types of home environment was also assessed by means of additional tests, including learning trials (Lansade et al. Reference Lansade, Valenchon, Foury, Neveux, Cole, Layé, Cardinaud, Lévy and Moisan2014; Greening et al. Reference Greening, Downing, Amiouny, Lekang and McBride2021), handling tests (Harewood & McGowan Reference Harewood and McGowan2005; Yarnell et al. Reference Yarnell, Hall, Royle and Walker2015), reactivity tests (Visser et al. Reference Visser, Ellis and Van Reenen2008) and ridden tests (Werhahn et al. Reference Werhahn, Hessel and Van den Weghe2012; Pessoa et al. Reference Pessoa, Trigo, Mesquita Neto, Lacreta Junior, Sousa, Muniz and Moura2016). Where the home environment was found to affect behaviour in these scenarios this is identified as a factor in the relevant section in Part II (Hall & Kay Reference Hall and Kay2024) of this review.

Table 2. (a) Behaviour in the home environment 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).

^a Postural attributes: Elevated neck with ears pricked forward, looking intensely at the environment; ^b Neck horizontal at same level as back, fixed stare, ears, and head static.

^c Agonistic behaviours: (chasing, contact biting, and kicking) and low-level threats (pinned ears, tail swishing, bite, and kick threats).

^d Positive response to allogrooming: neck moderately raised, eyes open or half-closed, upper lip extended, ears turned backwards almost in line with the nose.

Ethograms were used in the 21 retained articles. One study (Stomp et al. Reference Stomp, Leroux, Cellier, Henry, Lemasson and Hausberger2018a) focused on a very specific behaviour (snorts), and there were only two examples where lists of behaviour were used without further descriptors (Werhahn et al. Reference Werhahn, Hessel and Van den Weghe2012 for the tests of ridden behaviours; Lansade et al. Reference Lansade, Valenchon, Foury, Neveux, Cole, Layé, Cardinaud, Lévy and Moisan2014: response to enrichment). Where the ethogram had been developed with reference to previous studies, a total of 30 different references were cited as sources, with only three being cited by more than one study (McDonnell & Haviland Reference McDonnell and Haviland1995; Cooper et al. Reference Cooper, McDonald and Mills2000; McDonnell Reference McDonnell2003 were each cited by two studies). The behaviours recorded varied according to the focus of the study. Activity patterns and behaviour in situ were used to assess the impact of housing type (Visser et al. Reference Visser, Ellis and Van Reenen2008; Werhahn et al. Reference Werhahn, Hessel and Van den Weghe2012; Yarnell et al. Reference Yarnell, Hall, Royle and Walker2015; Pessoa et al. Reference Pessoa, Trigo, Mesquita Neto, Lacreta Junior, Sousa, Muniz and Moura2016; Ribeiro et al. Reference Ribeiro, Matzkeit, Nicolau JTde, Castilha, de Oliveira and Bankuti2019; Ruet et al. Reference Ruet, Lemarchand, Parias, Mach, Moisan, Foury, Briant and Lansade2019; Marliani et al. Reference Marliani, Sprocatti, Schiavoni, Bellodi and Accorsi2021), with studies assessing sleep patterns including detail of types of recumbency (Burla et al. Reference Burla, Rufener, Bachmann, Gygax, Patt and Hillmann2017; Greening et al. Reference Greening, Downing, Amiouny, Lekang and McBride2021). Only behaviours identified in previous studies as being indicative of stress were recorded by Harewood and McGowan (Reference Harewood and McGowan2005), but no specific references were cited to support the ethogram used. Stereotypical and other abnormal repetitive behaviours were recorded by Lesimple et al. (Reference Lesimple, Gautier, Benhajali, Rochais, Lunel, Bensaïd, Khalloufi, Henry and Hausberger2019), Ribeiro et al. (Reference Ribeiro, Matzkeit, Nicolau JTde, Castilha, de Oliveira and Bankuti2019), Ruet et al. (Reference Ruet, Lemarchand, Parias, Mach, Moisan, Foury, Briant and Lansade2019), Sénèque et al. (Reference Sénèque, Lesimple, Morisset and Hausberger2019) and Sauveroche et al. (Reference Sauveroche, Henriksson, Theodorsson, Svensson Holm and Roth2020). Sénèque et al. (Reference Sénèque, Lesimple, Morisset and Hausberger2019) also recorded the posture of the horse in situ.

When behaviour in horses being kept in groups was recorded, the focus was on the type of social interactions observed and the extent to which individual behaviour was affected by other group members (feeding, drinking, and resting behaviours). Negative social interactions, primarily levels of aggression and avoidance behaviours were recorded in relation to different approaches to forage provision (Melvin et al. Reference Melvin, Costello and Colpoys2020; Sundmann et al. 2022; Baumgartner et al. Reference Baumgartner, Erhard and Zeitler-Feicht2023) and around water resources (Hildebrandt et al. Reference Hildebrandt, Büttner, Krieter and Czycholl2022). Social disturbance of lying bouts was recorded by Burla et al. (Reference Burla, Rufener, Bachmann, Gygax, Patt and Hillmann2017) and by Hildebrandt et al. (Reference Hildebrandt, Büttner, Krieter and Czycholl2022). Both altruistic and different levels of agonistic behaviour were recorded by Suagee-Bedore et al. (Reference Suagee-Bedore, Linden and Bennett-Wimbush2021) in relation to the size of the group turn-out area. The effect of one form of environmental enrichment (auto-brushes) on group-housed horses was assessed in terms of social interactions (both positive and negative) and facial expression when using the brushes by Lansade et al. (Reference Lansade, Lemarchand, Reigner, Arnould and Bertin2022). See Table 2(b) for factors found to affect behaviour in the home environment.

Supporting evidence

The supporting evidence presented in the retained studies was variable. Reference to past studies and comparisons with physiological measures were most frequently used. Where horses were able to demonstrate choice or social interactions were recorded, approach behaviour and affiliative social interactions were deemed positive signs, while avoidance, and social disturbance/agonistic interactions were deemed negative signs (at least for the recipient). This attribution of affective state to different types of social interaction was supported by physiological measures in several studies. Suagee-Bedore et al. (Reference Suagee-Bedore, Linden and Bennett-Wimbush2021) found a reduction in both plasma cortisol levels and agonistic/mild threat behaviours in relation to increased enclosure size. Sauveroche et al. (Reference Sauveroche, Henriksson, Theodorsson, Svensson Holm and Roth2020) found that increased positive social behaviour and resting behaviour both correlated with decreased concentrations of hair cortisol. Although Baumgartner et al. (Reference Baumgartner, Erhard and Zeitler-Feicht2023) found both negative social interactions and salivary cortisol levels increased in the condition when there were fewer feeding stations, the authors noted that the level of physiological response was lower than in studies investigating the effect of other stressors (for example, transport). Similarly, Melvin et al. (Reference Melvin, Costello and Colpoys2020) found that although agonistic threats and avoidance behaviour varied in relation to how forage was presented, no difference in plasma cortisol levels was found.

The effect of individual compared to group housing on behaviour and physiology was assessed in both adult horses and in youngsters stabled for the first time. Visser et al. (Reference Visser, Ellis and Van Reenen2008) recorded the response of young horses to first-time stabling, either individually or in pairs. Individually housed horses displayed more behaviours indicative of stress (neighing, pawing, nibbling, snorting), spent more time standing vigilant and sleeping, and less time eating than those in paired housing. At the end of the 12-week trial period, 67% of the horses housed individually were seen performing one or more types of stereotypical behaviour. The response to a corticotropin-releasing factor (CRF) challenge test was reduced in those stabled individually, suggested by the authors to be the result of a desensitisation of the hypothalamic-pituitary-adrenal (HPA) axis in response to the stress associated with isolation (Visser et al. Reference Visser, Ellis and Van Reenen2008). Individually housed horses had higher mean heart rates in the stable than those in paired housing (Visser et al. Reference Visser, Ellis and Van Reenen2008). Harewood and McGowan (Reference Harewood and McGowan2005) recorded more stress-related behaviours when fillies were individually housed for the first time than during the previous 24 h when in groups in an outdoor paddock but found no differences in salivary cortisol levels or heart rate. Foals housed in an enriched environment that included turn-out time in groups showed less vigilant/aberrant behaviour, and increased time lying and with ears forward than foals kept in a non-enriched stable environment with only individual turn-out (Lansade et al. Reference Lansade, Valenchon, Foury, Neveux, Cole, Layé, Cardinaud, Lévy and Moisan2014). After eleven weeks almost 50% of the foals in the control environment did not finish their feed. After six weeks higher levels of salivary cortisol were found in the morning in the enriched group (Lansade et al. Reference Lansade, Valenchon, Foury, Neveux, Cole, Layé, Cardinaud, Lévy and Moisan2014). These three studies highlight the complexity associated with using cortisol concentration as a means of interpreting behaviour. Young horses showed behavioural signs of negative affective state following a short-term housing experience, but no physiological response was recorded (Harewood & McGowan Reference Harewood and McGowan2005). Longer-term exposure to a seemingly negative home environment was found to desensitise the HPA response by Visser et al. (Reference Visser, Ellis and Van Reenen2008) and Lansade et al. (Reference Lansade, Valenchon, Foury, Neveux, Cole, Layé, Cardinaud, Lévy and Moisan2014) found increased cortisol levels associated with an enriched environment.

The effect of age on cortisol responses and the extent to which this physiological measure can be used to assess the valence of affective state as opposed to arousal is currently unclear. In adult horses, higher levels of faecal corticosterone were found when they were housed individually (as compared with paired or group housing), but the only behavioural differences recorded were an increase in feeding and a decrease in standing behaviour when horses were paddock-housed in a group, which could be accounted for by the availability of grass (albeit relatively bare) and space in this condition (Yarnell et al. Reference Yarnell, Hall, Royle and Walker2015). Considerable individual differences were found when heart-rate variability (HRV) was monitored by Werhahn et al. (Reference Werhahn, Hessel and Van den Weghe2012), but in general, this was lower in horses with no turn-out time, with these horses spending less time lying down in the stable than those who were allowed turn-out. The latter finding could be explained by an increase in feeding time and decreased resting time during turn-out.

As an additional means of assessing how the home environment related to affective state, additional tests were included in six of the retained studies. Visser et al. (Reference Visser, Ellis and Van Reenen2008) found no difference in response to a novel object test between the housing groups. Harewood and McGowan (Reference Harewood and McGowan2005) found that fillies were less relaxed during saliva sampling after being housed individually for 24 h (compared to when group-housed in a paddock). Yarnell et al. (Reference Yarnell, Hall, Royle and Walker2015) found that individually housed horses were less relaxed during temperature sampling than when group or pair housed. Less aggression towards humans was found by Ruet et al. (Reference Ruet, Lemarchand, Parias, Mach, Moisan, Foury, Briant and Lansade2019) when horses were housed in stables with a window to the outside and on straw (as opposed to non-straw) bedding. During ridden work, rider-assessed ‘willingness to perform’ was scored lowest in those horses with no turn-out (Werhahn et al. Reference Werhahn, Hessel and Van den Weghe2012). Similarly, when horses stabled in total confinement were assessed by riders after ridden police patrol work, they were deemed to be more alert, nervous, and fearful than when allowed partial turn-out (Pessoa et al. Reference Pessoa, Trigo, Mesquita Neto, Lacreta Junior, Sousa, Muniz and Moura2016).

Limiting factors

The retained articles relating to the home environment varied greatly in terms of objectives, experimental design and behaviour recorded. In addition, justification for attributing affective state to specific behaviours relied primarily on past work, with some findings being supported by means of physiological measures and/or responses to handling and ridden activities. However, when recording behaviour in different situations it is important to highlight the fact that certain behaviours can only occur if resources are provided to facilitate them (for example, social behaviour, movement, feeding behaviour) and these limitations must be accounted for. Also, in most studies there was a bias towards recording behavioural signs of negative affective state, with less frequent conclusions relating to behavioural signs of positive affective state.

Identifying behaviour in situ in the home environment that is indicative of positive affective state is limited by and dependent upon features of that environment. Where individual stabling, lack of space and limited access to forage are features of the home environment, the behavioural repertoire of the horse is severely restricted. Whether or not the subjective experience of horses kept in such an environment can ever be positive is debatable and different interpretations of similar behaviours were evident in this review. For example, ‘alert posture (elevated neck with ears pricked forward)’ was deemed a positive sign of interest in the environment by Ruet et al. (Reference Ruet, Lemarchand, Parias, Mach, Moisan, Foury, Briant and Lansade2019) and negative, vigilant behaviour by Lansade et al. (Reference Lansade, Valenchon, Foury, Neveux, Cole, Layé, Cardinaud, Lévy and Moisan2014). A similar discrepancy was found in relation to whether increased time spent standing relaxed is a positive sign (Sauveroche et al. Reference Sauveroche, Henriksson, Theodorsson, Svensson Holm and Roth2020), or increased time standing dozing a negative one (Visser et al. Reference Visser, Ellis and Van Reenen2008) (see Table 2[a]). Such interpretations can both potentially be correct, as informed by the context, and the degree to which they occur, but this does highlight the challenge when attributing affective state to behavioural signs. An association between certain postural morphometric measurements taken when the horse was standing in the stable and whether the horse was reported as displaying stereotypical behaviour could inform differentiation between positive and negative signs (Sénèque et al. Reference Sénèque, Lesimple, Morisset and Hausberger2019), although potentially confounded by the type and conformation of individual horses. Behavioural signs of negative affective state were reported more consistently, and these included stereotypical behaviours (Visser et al. Reference Visser, Ellis and Van Reenen2008; Ribeiro et al. Reference Ribeiro, Matzkeit, Nicolau JTde, Castilha, de Oliveira and Bankuti2019) and behavioural events such as ears backwards, sniffing, licking stable walls/bars, repetitive head swinging (Lansade et al. Reference Lansade, Valenchon, Foury, Neveux, Cole, Layé, Cardinaud, Lévy and Moisan2014), pawing, stamping, kicking at walls, and high-pitched whinnies (Visser et al. Reference Visser, Ellis and Van Reenen2008; Lesimple et al. Reference Lesimple, Gautier, Benhajali, Rochais, Lunel, Bensaïd, Khalloufi, Henry and Hausberger2019), as shown in Table 2(a).

In summary

The results of the studies retained all support the conclusion that equine welfare is likely to be compromised by individual, restrictive housing. This was reflected in behaviour expressed in situ and in behavioural responses to handling and ridden tests (see also Part II; Hall & Kay Reference Hall and Kay2024). The need for company and the negative consequences of lack of social contact were demonstrated in young horses where reduced feeding time and subsequently reduced weight gain were reported when they were housed individually as opposed to in pairs (Visser et al. Reference Visser, Ellis and Van Reenen2008), and in adult horses in the study by Yarnell et al. (Reference Yarnell, Hall, Royle and Walker2015). Other features of the home environment found to be necessary for the horse to fulfil its behavioural needs, and by implication to impact on its affective state, included sufficient space and adequate forage provision. Specific factors reported in the retained studies are shown in Table 2(b).

The physical environment

The impact of physical features of the home environment on social interactions was assessed by Majecka and Klawe (Reference Majecka and Klawe2018) and Pierard et al. (Reference Pierard, McGreevy and Geers2019). An increase in both affiliative and agonistic interactions was associated with a decrease in enclosure size (Majecka & Klawe Reference Majecka and Klawe2018), and aggressive threats increased with increases in stocking density (Pierard et al. Reference Pierard, McGreevy and Geers2019). Affiliative interactions (allogrooming) between preferred partners increased in frequency but were of a decreased duration in confined (n = 85) as opposed to open pasture (n = 115) settings in large established groups of mares (Kieson et al. Reference Kieson, Goma and Radi2023). Other behavioural signs found to be associated with negative valence (agonistic behaviour: Luz et al. Reference Luz, Maia, Pantoja, Neto and Puoli Filho2015; discomfort: Torcivia & McDonnell Reference Torcivia and McDonnell2021), including increased locomotion and vigilant stances were observed in the confined condition (Kieson et al. Reference Kieson, Goma and Radi2023). In addition to spatial differences in the studies by Majecka and Klawe (Reference Majecka and Klawe2018) and Kieson et al. (Reference Kieson, Goma and Radi2023), the larger enclosures also provided more forage/grazing opportunities which are likely to have contributed to behavioural differences.

Group stability

Group instability and change resulted in an increase in agonistic interactions in young (two years of age) domestic mares (Christensen et al. Reference Christensen, Søndergaard, Thodberg and Halekoh2011) and physiological signs of stress in feral mares that had been treated with porcine zona pellucida (PZP) contraception (Nuñez et al. Reference Nuñez, Adelman, Smith, Gesquiere and Rubenstein2014). In the latter study, the conclusion that these group changes had a negative impact on the mares was informed by an increase in faecal cortisol metabolites associated with these group changes (Nuñez et al. Reference Nuñez, Adelman, Smith, Gesquiere and Rubenstein2014). In domestic horses, an increase in standing alert and decrease in time spent grazing was found when mares were removed from their group and either isolated or kept with a single companion for a short period (6 h). However, no effect on plasma cortisol level or heart rate was found when compared with mares that had been returned to their group rather than put into isolation (Strand et al. Reference Strand, Tiefenbacher, Haskell, Hosmer, McDonnell and Freeman2002). Although increases in whinnying, urination and rolling were recorded, there was no difference between mares kept alone or with a companion, and the authors note that there were three factors that could account for these behaviours: the novelty of the environment, isolation, and separation from the established herd (Strand et al. Reference Strand, Tiefenbacher, Haskell, Hosmer, McDonnell and Freeman2002). Whether this short-term separation from the established group resulted in a change in social interactions when the mares rejoined the group was not reported.

Group structure and composition

In feral horses, affiliative interactions (including grooming) were associated with a right-side preference which the authors (da Cruz et al. Reference da Cruz, Hirata, dos Santos and Mendonça2023) suggested was linked to the left-brain hemisphere dominance for positive affect and affiliative behaviours (Ahern & Schwartz Reference Ahern and Schwartz1979). This finding contrasts with that of Farmer et al. (Reference Farmer, Krüger, Byrne and Marr2018) where domestic horses were shown to have a left-side preference during allogrooming, a consequence perhaps of human handling which tends to occur more frequently from the left side of the horse. Differences between Przewalski (Equus ferus przewalskii) and domestic horses were found by Górecka-Bruzda et al. (Reference Górecka-Bruzda, Fureix, Ouvrard, Bourjade and Hausberger2016) in relation to the association between social interactions and yawning behaviour. There was a positive correlation between the frequency of yawning and aggressive interactions in Przewalski horses and between yawning frequency and affiliative interactions in domestic horses. It was suggested that yawning could relate to the intensity, rather than the valence, of a social interaction (Górecka-Bruzda et al. Reference Górecka-Bruzda, Fureix, Ouvrard, Bourjade and Hausberger2016).

As well as differences between feral and domestic horses, other characteristics have been shown to affect the type and frequency of social interactions. Groups comprising ponies (Shetland and Welsh) had more friendly interactions than mixed horse groups (Majecka & Klawe Reference Majecka and Klawe2018). Mares in oestrus were recorded as standing for longer and spending less time in locomotion than those in dioestrus, and an increase in the number of agonistic interactions occurred at the start of dioestrus (Stachurska et al. Reference Stachurska, Kędzierski, Kaczmarek, Wiśniewska, Żylińska and Janczarek2023). Less time was spent active, more time feeding, and fewer social interactions were recorded in lactating compared with non-lactating mares (York & Schulte Reference York and Schulte2014). Górecka-Bruzda et al. (Reference Górecka-Bruzda, Jastrzębska, Drewka, Nadolna, Becker and Lansade2022) found mares to be more socially dependent than castrated males. In a two-choice test between access to food or a companion, more mares than geldings chose the companion over food. Such choice tests provide valuable insights into factors that are important to the horses and are likely to promote positive affect.

Two studies included observations of play behaviour in young (two years of age: Christensen et al. Reference Christensen, Søndergaard, Thodberg and Halekoh2011) and adult (Hausberger et al. Reference Hausberger, Fureix, Bourjade, Wessel-Robert and Richard-Yris2012) horses. Play behaviour varied between different groups of young horses and was even more variable in unstable as opposed to stable groups (Christensen et al. Reference Christensen, Søndergaard, Thodberg and Halekoh2011). In adult horses, those exhibiting play behaviour were found to be more likely to also have potentially painful vertebral issues, and showed more aggression towards humans, than non-players (Hausberger et al. Reference Hausberger, Fureix, Bourjade, Wessel-Robert and Richard-Yris2012), but further evidence is required before wider conclusions can be drawn in relation to underlying affective state.

Communication of affective state

The communication of affective state by means of vocalisation was reported in six retained studies (Pond et al. Reference Pond, Darre, Scheifele and Browning2010; Briefer et al. Reference Briefer, Maigrot, Mandel, Freymond, Bachmann and Hillmann2015, Reference Briefer, Mandel, Maigrot, Briefer Freymond, Bachmann and Hillmann2017; Maigrot et al. Reference Maigrot, Hillmann, Anne and Briefer2017, Reference Maigrot, Hillmann and Briefer2022; Stomp et al. Reference Stomp, Leroux, Cellier, Henry, Hausberger and Lemasson2018b). Differences in vocalisation (whinnies) according to the valence of the situation were reported by Pond et al. (Reference Pond, Darre, Scheifele and Browning2010), and Briefer et al. (Reference Briefer, Maigrot, Mandel, Freymond, Bachmann and Hillmann2015) and a specific vocalisation (snort) was reported by Stomp et al. (Reference Stomp, Leroux, Cellier, Henry, Hausberger and Lemasson2018b) to be associated with positive experiences. Differences in the frequencies, energy spectrum and duration of horse vocalisations have been shown to be associated with the valence and arousal level of the underlying emotion, with physiological responses (HRV, RR) reliably changing with arousal level but not valence (Briefer et al. Reference Briefer, Maigrot, Mandel, Freymond, Bachmann and Hillmann2015). A weak (but inconsistent) association between both arousal and the valence conveyed within the vocalisation and changes in skin temperature (ST) was however found (Briefer et al. Reference Briefer, Maigrot, Mandel, Freymond, Bachmann and Hillmann2015). Species-specific differences may occur in the way that valence is communicated within vocalisations (Maigrot et al. Reference Maigrot, Hillmann, Anne and Briefer2017) but some cross-species recognition of both valence and arousal communicated through vocalisation has been demonstrated in the horse (Maigrot et al. Reference Maigrot, Hillmann and Briefer2022).

Valence has been associated with specific features of horse facial expression (Wathan et al. Reference Wathan, Burrows, Waller and McComb2015), and when presented with 2D images of horse faces portraying positive or relaxed expressions, or agonistic expressions, chose to approach the former and avoid the latter (Wathan et al. Reference Wathan, Proops, Grounds and McComb2016). Horses clearly use both auditory and visual information generated by conspecifics to guide social behaviour, facilitate positive interactions, and avoid aggressive encounters.

In summary

The physical and social environment combine to determine whether the species-specific needs of the horse can be fulfilled. There is a consensus that group membership is a priority for the horse and that the valence of social interactions contributes to and is evidence of individual affective state. Environmental conditions, group structure and composition, and individual characteristics have all been shown to relate to the type and frequency of social interactions occurring within that group, which in turn provide an insight into the affective state of group members. Behavioural signs of group cohesion and the nature of within-group social interactions reflect the quality of life of horses within a social group. Although short-term separation from the group was not found to have a negative impact on some individuals, it is not clear how such removals affect the overall stability of the group. In addition, behaviours associated with separation anxiety can cause safety issues for human handlers as well as being signs of distress in the horse (see Hartmann et al. Reference Hartmann, Christensen and Keeling2011). Regardless, providing the horse with companionship in the form of a stable group of conspecifics is imperative to satisfy social behavioural needs. The potentially negative effect of removal from the group may be lessened by repeated short-term experiences, followed consistently by return to the group. Further evidence is required to determine how the negative impact of separation from the group can be reduced, but keeping horses alone is not an acceptable solution. Factors found to affect behavioural signs of affective state in social situations are shown in Table 3(b).

Pre-weaning mare-foal separation

The effect of short-term (10-min) physical but not auditory separation of mare and foal at two, four, six, eight, ten and twelve weeks of age, on maternal care and subsequent response to weaning was assessed by Moons et al. (Reference Moons, Laughlin and Zanella2005). Post short-term separation both mare and foal looked at each other more frequently and the foal initiated physical contact with the mare more frequently than pre-separation. At weaning there was no difference in HR response between foals that had been separated (n = 5) and the control group (n = 5), but the latter showed more locomotion. The foals that had been exposed to short-term separation had higher concentrations of salivary cortisol on the post-weaning morning (Moons et al. Reference Moons, Laughlin and Zanella2005), possibly a response to the failure to be reunited after this separation. Consistent individual variation in the response of foals to pre-weaning separation from the mare (for 2 min) as tested four times from one week post-natal to six months was found by Pérez Manrique et al. (Reference Pérez Manrique, Hudson, Bánszegi and Szenczi2019) in both behaviour (vocalisation rate, raised tail and locomotion) and physiology (HRV). Due to the small sample size and contradictory behavioural and physiological findings by Moons et al. (Reference Moons, Laughlin and Zanella2005) and the individual variation in foal response to maternal separation, with no subsequent measures taken following weaning (Pérez Manrique et al. Reference Pérez Manrique, Hudson, Bánszegi and Szenczi2019), there is no conclusive evidence that pre-weaning separation reduces the stress of the final physical separation.

A two-stage approach to weaning was trialled by Merkies et al. (Reference Merkies, DuBois, Marshall, Parois, Graham and Haley2016) with nutritional separation (udder cover fitted to the mare to prevent nursing) for four days prior to physical separation. During the period of nutritional separation, foals were found to spend more time lying down than controls, but no behavioural or physiological differences between groups were found after physical separation. The only difference found between the groups was in the mares, where nutritional separation resulted in less vocalisation following physical separation. After physical separation, all foals were recorded as vocalising more, were more active and aggressive, and spent less time lying down. Increases in faecal cortisol metabolites were recorded in all mares and foals following physical separation, but individual variation in both behavioural and physiological responses was again found (Merkies et al. Reference Merkies, DuBois, Marshall, Parois, Graham and Haley2016). A gradual increase in the time spent physically separated (by a fence that allowed mares and foals visual, olfactory, and tactile contact, but prevented suckling) over the month preceding weaning was shown to reduce time spent trotting and vocalising compared with foals weaned without this preparation, as well as resulting in reduced concentrations of salivary cortisol (Lansade et al. Reference Lansade, Foury, Reigner, Vidament, Guettier, Bouvet, Soulet, Parias, Ruet, Mach, Lévy and Moisan2018). This gradual weaning process also benefited the mares who showed fewer behavioural signs of stress and lower salivary cortisol levels than mares whose foals had been weaned without this progressive separation (Lansade et al. Reference Lansade, Foury, Reigner, Vidament, Guettier, Bouvet, Soulet, Parias, Ruet, Mach, Lévy and Moisan2018).

Physical and social environment

The physical and social environment has been shown to be one key factor in how foals cope with artificial weaning and in their long-term well-being. For example, the development of stereotypical behaviour has been associated with confinement and isolation of foals during the weaning process (Waters et al. Reference Waters, Nicol and French2002). Regardless of how weaning is managed, changes in foal behaviour include increased vocalisation, locomotion, and reduced feeding, but these can be less pronounced if foals remain in a familiar environment and familiar group with known adult (unrelated) mares, compared with foals kept in peer groups without an adult (Erber et al. Reference Erber, Wulf, Rose-Meierhöfer, Becker-Birck, Möstl, Aurich, Hoffmann and Aurich2012). Foals moved to an unfamiliar environment at weaning, with some unfamiliar group members (foals) were found to spend an increased time lying down initially, followed by resting standing for longer periods than before they were weaned (Delank et al. Reference Delank, Reese, Erhard and Wöhr2023). A similar impact of an unfamiliar stable environment on general apathy in foals at weaning was found by Górecka-Bruzda et al. (Reference Górecka-Bruzda, Suwała, Palme, Jaworski, Jastrzebska, Boroń and Jezierski2015), with an increased time spent standing, decreased social activity and reduced feeding recorded in forest-reared compared with stable-reared foals.

In contrast with the findings of Visser et al. (Reference Visser, Ellis and Van Reenen2008) (see The home environment), Hoffman et al. (Reference Hoffman, Kronfel, Holland and Greiwe-Crandell1995) recorded fewer behavioural and physiological signs of stress in foals housed individually at weaning than in those housed in pairs. Foals housed in pairs stood still less than those housed individually (deemed a negative sign by Delank et al. Reference Delank, Reese, Erhard and Wöhr2023), with a high frequency of aggressive interactions recorded (Hoffman et al. Reference Hoffman, Kronfel, Holland and Greiwe-Crandell1995). The increased cortisol response in pair-housed foals may have related to the increased arousal associated with social interactions, regardless of valence. Lack of social contact during periods of isolation was found to reduce outward behavioural signs of emotional reactivity in yearlings (Lansade et al. Reference Lansade, Neveux and Levy2012, see Part II; Hall & Kay Reference Hall and Kay2024). The association between having a companion (or not) and the extent to which emotion is communicated behaviourally may explain in part the findings of Hoffman et al. (Reference Hoffman, Kronfel, Holland and Greiwe-Crandell1995). Also, Henry et al. (Reference Henry, Zanella, Sankey, Richard-Yris, Marko and Hausberger2012) found that foals kept in groups that included unrelated (and in this case, unfamiliar) adult horses, rather than in peer-only groups vocalised less and had lower salivary cortisol responses. Only foals kept in groups without an adult horse showed increased aggression towards peers and abnormal behaviour such as wood-chewing (Henry et al. Reference Henry, Zanella, Sankey, Richard-Yris, Marko and Hausberger2012).

Sex-related differences

Although there has been shown to be individual variation in the responses of mares and foals, artificial weaning is always associated with both behavioural and physiological signs of stress. Some differences between fillies and colts in both behaviour and physiological responses were found. Colts spent more time eating than fillies (Hoffman et al. Reference Hoffman, Kronfel, Holland and Greiwe-Crandell1995), and on the day of weaning walked less and defaecated more than fillies (Moons et al. Reference Moons, Laughlin and Zanella2005; Górecka-Bruzda et al. Reference Górecka-Bruzda, Suwała, Palme, Jaworski, Jastrzebska, Boroń and Jezierski2015; Wulf et al. Reference Wulf, Beythien, Ille, Aurich and Aurich2018). Subsequently, colts demonstrated more curiosity about the environment, were more socially interactive and lay down less often than fillies (Górecka-Bruzda et al. Reference Górecka-Bruzda, Suwała, Palme, Jaworski, Jastrzebska, Boroń and Jezierski2015). Although initial weight loss was found to be greater in fillies than colts by Wulf et al. (Reference Wulf, Beythien, Ille, Aurich and Aurich2018), contrary to other studies, they did not find any sex-related differences in feeding or lying behaviour. However, levels of salivary cortisol (Wulf et al. Reference Wulf, Beythien, Ille, Aurich and Aurich2018) and faecal cortisol metabolites (Górecka-Bruzda et al. Reference Górecka-Bruzda, Suwała, Palme, Jaworski, Jastrzebska, Boroń and Jezierski2015) were both higher in filly than colt foals at weaning. Despite many factors varying between studies, and some findings being based on relatively small sample sizes, there is evidence that there are sex-related differences in how foals respond to artificial weaning.

In summary

The behavioural and physiological response to artificial weaning cannot be avoided but the process can be made less onerous by adopting protocols that have been shown in the retained studies to lessen the negative impact of physical separation. The gradual approach to mare and foal separation outlined by Lansade et al. (Reference Lansade, Foury, Reigner, Vidament, Guettier, Bouvet, Soulet, Parias, Ruet, Mach, Lévy and Moisan2018) had much to recommend it. A priority should be to ensure that the weanling foals are kept in social groups that are familiar to them, including familiar adults, and in an environment that is also familiar to them. Although potentially challenging to many horse breeders, the short- and long-term well-being of these youngsters must be the primary concern. In agreement with Górecka-Bruzda et al. (Reference Górecka-Bruzda, Suwała, Palme, Jaworski, Jastrzebska, Boroń and Jezierski2015), feeding behaviour is an indication of how foals are coping with separation from the mare, and potentially how mares themselves are coping. It is likely that mares will not be as adversely affected by environmental changes, but the provision of a stable social group and the freedom to move and feed are important factors that are likely to reduce the negative impact of separation from their foals. Factors affecting behaviour associated with weaning are shown in Table 4(b).

Limitations of this approach

Although judgement bias testing offers a degree of insight into the affective state of animals, it requires prior training (often a lengthy process), and performance has been found to be affected by previous approaches to training. The effect of training using positive or negative reinforcement on judgement bias was assessed by Briefer-Freymond et al. (Reference Briefer Freymond, Briefer, Zollinger, Gindrat-von Allmen, Wyss and Bachmann2014), with the finding that those previously trained using negative reinforcement made more positive appraisals of ambiguous stimuli than those trained using positive reinforcement. The food reward used in the judgement bias testing may have constituted a change for the better for those that had previously been negatively reinforced (Briefer-Freymond et al. Reference Briefer Freymond, Briefer, Zollinger, Gindrat-von Allmen, Wyss and Bachmann2014).

The search for more immediate objective measures of affective state has included comparisons with performance in judgement bias tests. A study evaluating eye wrinkles in horses during a judgement bias test found no significant association between these and evidence of optimism/pessimism based on performance (Hintze & Schanz Reference Hintze and Schanz2021). However, Marr et al. (Reference Marr, Farmer and Krüger2018) found a significant relationship between motor laterality and performance in judgement bias tests, with a positive judgement bias being displayed by horses that moved off using the right forelimb first. In the study by Marliani et al. (Reference Marliani, Vannucchi, Kiumurgis and Accorsi2022b), horses that had been trained that the rewarded stimulus position was on the right were found to approach the ambiguous stimulus faster than those horses trained to reward from the stimulus position on the left. In terms of more general applicability, such approach behaviour in non-experimental situations could be used as an indication of positive expectations, potentially linked to a positive mental state. However, whether lateral biases or facial expression can provide any further insight is yet to be determined.

In summary

The results of the retained studies involving judgement bias testing provide some further evidence that housing and management that fulfils the need for space and company is associated with a more optimistic (positive) appraisal of ambiguous stimuli by the horse. Although there were confounding factors identified, and a lack of correlation with other measures, studies did demonstrate that change for the better (in terms of addressing the behavioural needs of the horse) was associated with optimism and potentially positive affective state. See Table S1.4 in the Supplementary material for further details of the studies retained.

Factors relating to the occurrence of stereotypical behaviour

Arena et al. (2021) compared the current housing and management of horses reported by a veterinarian as displaying behavioural pathologies (oral stereotypies, locomotor stereotypies, social problems, and training behaviours), with those not displaying such behaviour (and deemed healthy). A positive correlation between the occurrence of behavioural pathologies, time spent in the stable and frequency of work, and a low-fibre, high-energy diet was found. Justification for deeming such behavioural pathologies to be signs of chronic stress and potentially negative affective state was based both on the findings of previous studies and on higher plasma cortisol levels compared with healthy horses. No difference in hair cortisol levels between horses displaying behavioural pathologies and their healthy counterparts was found (Arena et al. Reference Arena, Marliani, Sabioni, Gabai, Bucci and Accorsi2021).

It has been suggested that the type of work/role of the horse affects the prevalence and type of behaviour exhibited. Zuluaga et al. (Reference Zuluaga, Mira, Sánchez and Martínez2018) found behavioural signs of chronic stress (including wood chewing, pawing, bed eating, crib-biting, and windsucking) in urban police patrol horses in Columbia, but it was not clear whether this was a direct consequence of the role or management factors. Although differences between horses involved in different sporting disciplines (including dressage, showjumping, eventing, and endurance) have been found (in some cases independently of management differences, for example, Hausberger et al. Reference Hausberger, Gautier, Biquand, Lunel and Jégo2009), it is not clear whether these differences result from initial selection for a role or because of the demands of that role. Hanis et al. (Reference Hanis, Chung, Kamalludin and Idrus2021) found differences between horse working groups (leisure, endurance, patrol horses) in the prevalence of abnormal/stereotypic behaviour in the stable, but also related differences in dietary intake between groups, and concluded that this was a major factor in the development of such behaviour. In general, the effect of work/role on behaviour is confounded by many factors (including selection, training, and management) and findings should consequently be carefully evaluated in terms of all contributory variables. However, as confirmed by the findings of a recent survey by Kádár et al. (Reference Kádár, Maros, Drégelyi, Szedenik, Lukácsi, Pesti, Besenyei and Egri2023), the prime factor is the type of home environment, with an increased occurrence of abnormal repetitive behaviours reported in stabled horses as compared with those kept at pasture.

The association between stereotypies and other features of behaviour

The impact of performing stereotypical behaviours on learning performance has been evaluated. No effect of stereotypical behaviour on learning ability has been found (Briefer Freymond et al. Reference Briefer Freymond, Ruet, Grivaz, Fuentes, Zuberbühler, Bachmann and Briefer2019, Reference Briefer Freymond, Beuret, Ruet, Zuberbühler, Bachmann and Briefer2020), with the proviso that the performance of the stereotypical behaviour is not restricted. Horses that displayed crib-biting during a spatial learning task showed fewer behavioural signs of frustration and had lower levels of salivary cortisol than stereotypical horses that did not display this behaviour (Briefer Freymond et al. Reference Briefer Freymond, Beuret, Ruet, Zuberbühler, Bachmann and Briefer2020). An increased sensitivity to tactile stimuli was found in crib-biting horses, which the authors suggested may be linked to physiological responses to chronic stress, that also potentiated the development of this behaviour (Freymond et al. Reference Freymond, Bardou, Beuret, Bachmann, Zuberbühler and Briefer2019).

In summary

Regardless of the potential positive or negative impact that performing abnormal repetitive behaviours has on an individual’s QOL, research to date provides conclusive evidence that it is the failure to provide for the behavioural needs of the horse that is the causative factor. Although it is not clear whether the performance of stereotypies per se is a sign of current QOL, or a coping response resulting from past sub-optimal conditions (Sarrafchi & Blokhuis Reference Sarrafchi and Blokhuis2013), such behaviour has been shown to develop in young horses exposed to inappropriate housing and/or diet and persists throughout adulthood. As such, this behaviour provides evidence of the importance of providing for the behavioural needs of the horse throughout their life. Despite this, many horses continue to be managed in a way that fails to do so. See Table S1.5 in the Supplementary material for further details of the studies.