The survey received 185 participant responses in addition to the six expert responses. All experts were located in the U.K., as were 79.5% of the participants. Other participants were located in the U.S.A. (6.5%), Canada (4.9%), other European countries (4.9%), Australia (3.2%), South Africa (0.5%) and Brazil (0.5%). The preferred equestrian activities of participants were numbered as 28.1% dressage, 27.6% “happy hacking”, 10.8% eventing, 10.8% clicker training, 5.4% natural horsemanship, 4.3% endurance, 2.2% Western riding, 2.2% show jumping and 0.5% polo. A further 8.1% said that that they were “none of the above but enjoyed spending time with horses”; for convenience, this group was labelled “others” in later discussion. When asked about their experience with horses, 137 participants claimed to be experienced horse owners, of whom 20 were also professionals within the equine industry; 8 were experienced former owners of whom 2 were also professionals; 19 had owned horses for fewer than 5 years, 2 of whom were professionals; and 5 were not owners at all. A further 16 selected only the ‘professionals’ category. Of the 40 stating that they were professionals within the equine industry, 29 were instructors or trainers, with some also classifying themselves as behaviourists. In response to the question about ability to recognise behavioural expressions of fear, stress and anxiety, 75 claimed “yes, definitely”, 91 “yes, somewhat”, 18 “perhaps/sometimes” and 1 “probably not”.
3.1. Expert Responses
When asked generally to state the typical signs that they would use to determine whether a horse was frightened and/or stressed, the experts listed a variety of behavioural responses, of which a subset would likely be shown by an individual horse. While some were overt behaviours, such as rearing or flight, the majority were subtle, such as eye movements, breathing rate or yawning. Some comments reflected a wider understanding of causal factors such as avoidance of tack or changes in social behaviour. While detailed analysis of these qualitative responses is beyond the scope of this paper and will be the subject of a future work, the key point is that the experts were aware of potential early behavioural indications that are likely to reflect the onset of negative affect.
shows the responses given by the experts to the videos. The 13 affective states have been listed according to how many experts selected that option: for each video clip, the top line lists the responses selected by 4 or more experts, the bottom line lists unanimously rejected responses and the middle line lists the remaining responses. For example, the experts were unanimous in observing that the horse in video 1 exhibited behaviours associated with stress and did not exhibit behaviours associated with anger, enjoyment, excitement, play, relaxation or submissiveness. For each video, there was unanimity with respect to some video/affect pairs and disagreement for others. However, any contradiction tended to reflect the similar nature of the affective states (e.g., fear vs. anxiety) rather than any fundamental disagreement over the horse’s general experience.
3.2. Participant Responses
When asked to list the generic behavioural signs that they would use to determine whether a horse was frightened and/or stressed, many respondents were able to provide a set of subtle behaviours similar to those noted by the experts. However, a greater proportion of overt behaviours were also included, such as bolting, pushiness, vocalising, “foaming at the mouth” or trembling. While these would all have been correct responses, anecdotal experience of working in behavioural practice indicates that earlier and more subtle suggestions of distress would typically also have been present, consistent with the hierarchy of behavioural indicators of stress found in [9
All participants selected at least one affective state for each video clip and typically selected more than one. Figure 2
shows all affective state selections for each video. It was clear that the responses included a high number of “anxious”, “conflicted”, “fearful”, “frustrated”, “stressed” and “submissive” selections, in keeping with the experts. However, for some videos there was also a lower but consistent selection of those affective states that were typically not selected by the experts—“angry”, “enjoying it”, “excited”, “playful”, “relaxed” and “stubborn”. Such selections prompted the need for Criterion 2 in defining correct/incorrect responses. For example, a participant selecting “frustrated” and “fearful” for video 1 would be considered correct, whereas “frustrated” and “playful” would be considered incorrect.
a shows the total numbers of correct and incorrect participant responses as compared to the expert responses. There is a significant difference between the videos (
= 229.39, p
< 0.001; Cramer’s V = 0.46, p
< 0.001) with videos 1, 3 and 5 receiving significantly more correct answers, videos 2 and 4 receiving significantly more incorrect answers; video 6 was almost equal, receiving just one more correct answer than incorrect. Figure 3
b shows the total incorrect answers, according to whether they failed on account of Criterion 1, 2 or both. Again, the proportions depended on the video in question, with Criterion 1 prevalent in videos 2 and 5 and Criterion 2 prevalent in videos 1 and 3.
Pearson Chi-square tests also examined the associations between obtaining a correct answer and the participants’ age, preferred equestrian activity, experience and self-perceived ability to recognise equine distress. Fisher’s exact test was included to account for small sample sizes in some pairings. Videos 3 and 5 showed a significant association between increasing age and correct response, albeit with a relatively small effect size (
and 10.11, respectively, p
< 0.05; Cramer’s V = 0.27 and 0.24 respectively, p
< 0.05). Videos 1, 2 and 6 showed a significant association and medium effect size between preferred activity and correct response (
, respectively, p
< 0.05; Cramer’s V = 0.36, 0.39, 0.33, respectively, p
< 0.05). These results are illustrated in Figure 4
. Ignoring activities with fewer than five participants, video 1 was over-represented by “happy hackers”, dressage riders and endurance riders, video 2 by clicker trainers and “others” and video 6 by clicker trainers, endurance riders and “others”. Videos 3 and 6 showed a significant association between experience and correct response (
, respectively, p
< 0.05; Cramer’s V = 0.30, 0.26, respectively, p
< 0.05), but somewhat erratically, with experienced and new owners obtaining more correct responses in video 3 and professionals scoring higher in video 6. There was no association between correct response and self-reported ability to recognise fear, stress and anxiety for any of the videos.
Binary logistic regression analysis was used to explore the possible association between preferred equestrian activity, participant experience and successful recognition of negative affective states. A model with three predictor variables—Video, Activity and Experience—was used to predict the likelihood of correct answers (N
= 1,110). The model was successful in predicting 72.1% of the outcomes (p
< 0.01) compared with the null, intercept-only model. Table 2
lists the categories that had a significant effect on the model. The video categories made a significant contribution to the model, with an increasing likelihood (exp (B) > 1) of correct answers for videos 1, 3 and 5 and a decreasing likelihood (exp (B) < 1) for videos 2 and 4. Past experience and non-ownership made decreasing contributions to the model. Being experienced or a professional did not make a significant contribution to the likelihood of obtaining correct answers. Finally, the only significant (almost, p
= 0.05) contribution from the activity categories was clicker training, albeit with a very large upper confidence limit.
Finally, the participants were asked whether they would be happy for their own horse to be handled in a similar manner as observed in each video. Of particular interest was whether those who believed that less desirable affective states (angry, anxious, conflicted, fearful, frustrated, stressed, stubborn, submissive and switched-off) were present, and that the more pleasurable states (enjoying it, excited, playful and relaxed) were absent, would consider it an acceptable state for the horse to be in during training. The answers are represented in Figure 5
; total numbers of participants featured were n
= 171, 78, 160, 127, 144, 103 for each video, respectively. Black indicates those participants who, despite recognising negative affective states, would still permit their horses to be treated as in the video. Dark grey indicates a middle category for those who would “partially” permit the handling and light grey indicates those who would not. Of the sample of 185, just 6 participants were in the light grey category—i.e. considered the horse to be experiencing negative affective states and that it was unacceptable—for all 6 videos and 30 for videos 1–5. Of the 30, 11 were clicker trainers, compared with 6 claiming no particular activity, 4 dressage, 4 “happy hackers”, 2 eventers, 2 Western riders and 1 natural horsemanship trainer. Of the 30 who considered that the horses in videos 1–5 were experiencing negative affect and that it was unacceptable, 21 classified themselves as current or former experienced owners and 10 were professionals. Note that the complete sample included 40 professionals, 30 of whom, therefore, failed to recognise negative affective states or condoned the handling.