Here we investigated specific correlates of oral and motor stereotypic behaviours and glucocorticoid levels in two large and different samples of domestic horses, kept in sub-optimal conditions and already known to experience poor welfare states. We used plasma and also non-invasive faecal sampling to measure cortisol levels. According to the coping hypothesis of stereotypic behaviours – i.e. that stereotypic behaviours are performed as a means of helping the animal to cope with sub-optimal environments by reducing stress – horses that displayed stereotypic behaviours were expected to have lower cortisol levels than non-stereotypic horses. Furthermore within the sample of stereotypic horses, those with higher levels of stereotypic behaviours were expected to have lower cortisol levels. Both pC (study 1) nor fCM (study 1 and 2) concentrations however differed between stereotypic and non-stereotypic horses, nor were they significantly predicted by stereotypic behaviour levels, even when oral and motor stereotypic behaviour levels were considered separately. Therefore, cortisol measures do not indicate that stereotypic horses cope better, at least in terms of adrenocortical activity.
Our results confirm the earlier findings of Pell and McGreevy , Clegg et al.  and Hemmann et al.  indicating no significant relationship between plasma/salivary cortisol levels and stereotypic behaviours in this species. Interestingly, this absence of a relationship between plasma cortisol and stereotypic behaviour levels appears to be independent of the time of plasma sampling. Indeed, Hemmann and collaborators  report on not significant effects of crib-biting on cortisol circadian secretion when collecting plasma every two hours for 24 h in stereotypic and control non-stereotypic horses. When added to Pell and McGreevy’s  and Clegg’s et al.  previous studies, where plasma was collected in the morning, our complementary results based on plasma samples collected in the afternoon support Hemmann and collaborators’ conclusions. Moreover, in the four previous studies [7, 8, 12, 13] in which plasma samples were all collected in the morning, results were also contradictory, indicating that time of sampling cannot be the main explanation for results discrepancy. In this context, using faecal samples, that reflect an average level of circulating cortisol over a longer period, appears nevertheless to be a good complementary strategy.
Using faecal samples however might also raise a new methodological issue, namely a potential impact of variations in gut flora on the concentrations of cortisol metabolites. Indeed, the metabolites of cortisol that are measured in faeces are the products of extensive modification by bacteria in the gut (e.g. [35, 36]). As a consequence, the composition of bacteria could influence the type and/or the quantity of hormone metabolites (discussed in ). Imbalance in hindgut flora (as a result of acidosis) has been reported in horses displaying crib-biting  and one could expect this imbalance to influence fCM concentrations in these animals compared with non-stereotypic horses and to bias the results. FCM concentrations however did not differ between stereotypic and non-stereotypic horses, nor were they significantly predicted by stereotypic behaviour levels, even when oral and also motor stereotypic behaviour levels were considered separately. To our knowledge, motor stereotypic behaviours have not been reported to be linked with gastric nor gut disorders in horses; therefore horses displaying motor stereotypic behaviours would not be expected to differ from control horses in regards to their gut flora composition. In addition, no significant relationship between stereotypic behaviours and cortisol levels appeared either when taking blood samples measuring the actual hormone (a complementary approach advised by ). Therefore, our results that cortisol measures do not seem to indicate that stereotypic horses cope better appear unlikely to reflect methodological bias due to sampling methods, though further research on the potential effect of gut floral on fCM concentration is warranted.
The idea that stereotypic behaviours may help animals to cope with sub-optimal environments is not new e.g.[39–41], but it is still a highly debated issue and evidence is sparse and contradictory. McBride and Cuddeford  report higher plasma concentrations in horses immediately prior to the onset of a crib-biting bout, followed by a significant reduction post- crib-biting, suggesting that stereotypic behaviours may have a coping function that reduces stress levels in the animal. According to these results, the prevention of stereotypic behaviours may then be even more stressful and should induce a rise in glucocorticoids. In order to further test this prediction, two experimental studies have examined the effects of stereotypic behaviour prevention under controlled conditions on cortisol levels in horses. However, the environmental modifications used to prevent the stereotypic behaviours (i.e. devices such as crib-strap and anti-weave bar, removal of both a cribbing bar and hay) induced a rise in plasma cortisol concentrations in stereotypic horses, but also in control non-stereotypic horses [6, 12]. Thus, the observed rise in glucocorticoids in stereotypic horses, initially expected to reflect a stress response due to the prevention of stereotypic behaviours but also observed in non-stereotypic animals, might simply reflect the horse’s response to environmental changes as previously suggested in rodents e.g..
Note also that, as reported in the literature, differences in cortisol levels between stereotypic and non-stereotypic horses do not support the coping hypothesis prediction. Indeed, McGreevy and Nicol  and Bachmann et al.  reported higher plasma cortisol concentrations in stereotypic horses than in non-stereotypic horses. One however may argue that higher cortisol levels in stereotypic animals could represent methodological bias, particularly when moving horses from their home stable to an experimental one and/or using frequently repeated blood sampling, which both could be stressful for the animals [27, 28]. Stereotypic horses may be particularly stress-sensitive individuals and they may perceive environmental changes and blood sampling as even more stressful than non-stereotypic animals, which may have impaired the results. In this context, using faecal samples – i.e. a totally non-invasive measurement, yielding no bias due to sampling procedure – to assess adrenocortical activity appears to be a better strategy and reinforces the earlier findings cited above that no relationship is evidenced between stereotypic behaviours and basal glucocorticoid levels.
On the whole, our results also do not support the coping hypothesis prediction, though several explanations for these negative results can be proposed. The absence of difference between stereotypic and non-stereotypic horses might reflect equally low cortisol levels, an interpretation which would, at least partly, support a coping function of stereotypic behaviours. Bearing in mind however that horses can express poor welfare states in various ways (e.g. aggressiveness , apathy and unresponsiveness [19, 43]), one may also argue that, in the experienced sub-optimal life conditions, both stereotypic and non-stereotypic horses were equally stressed, i.e. equally high cortisol levels.
Another possible explanation for the results is that chronic stress levels might not be accurately measured by basal cortisol concentrations e.g.[34, 44]. Cortisol has a circadian rhythm (highest in the early morning, lowest in the evening), a phenomenon well-evidenced in horses [45–50]. Some chronic stressors have been reported to induce dysregulated pattern of hormone secretion, e.g. a flattened diurnal rhythm . Highly frequent blood collection can be a disturbing procedure in itself and is also clearly not always practicable or possible, especially in this study performed on working horses from riding schools. Including a pC concentration assessment at least at the times where cortisol concentrations are highest and lowest (the initial protocol of study 1, but not taken due to methodological reasons, see “methods”) would allow a calculation of a slope as an indication of cortisol’s rhythm across the day, and would provide additional interesting information. Note, however, that the effect of chronic stress on glucocorticoid circadian variations is still far from straightforward, as it might vary according to the stressor and individual features . Thus it would also be interesting to add other measures of coping to test further the coping hypothesis prediction.
Beyond the potential functional explanation of stereotypic behaviours proposed by the coping hypothesis, another explanation for our negative findings may be that the performance of stereotypic behaviour is not associated with stress or coping at all. Indeed, some forms of stereotypic behaviours are preservative and animals cannot stop performing them ; stereotypic behaviours can become habit forming (e.g. enhancement of habit formation in crib-biting horses ) and thus does not result in changes in cortisol levels; and/or the eliciting stimuli might no longer be present in the current environment e.g.. Note however that the latter issue would be hard to test in our population since these horses are still experiencing several challenges to their welfare, such as confinement, social isolation and time-restricted feeding practices.