The main aim of this study was to assess whether the addition of regular positive reinforcement training to horses’ weekly training and management has an effect on their optimism, stress levels and the horse-human relationship. I found that the treatment group increased their contact-seeking behaviour towards an unfamiliar person and showed fewer stress-related behaviours during a motionless human test after eight to nine weeks of positive reinforcement training, but there was no significant difference in their optimism as assessed by a cognitive bias test. 

Horses in the treatment group spent significantly more time engaging in physical contact with the unfamiliar person after the treatment period compared to before, while the control horses did not. These results support previous studies that show that horses trained with positive reinforcement display more contact-seeking behaviour towards strangers.

The treatment horses also showed a significant decrease in stress-related behaviours after the treatment period. These findings also align with previous studies that found lower stress levels in horses trained with positive reinforcement.

One of the stress-related behaviours in the ethogram was headshaking, and treatment horses showed significantly more headshaking before the treatment period than after. This higher frequency of headshaking could be confounded by late-summer flies in September. However, this fly effect should be seen in both treatment and control horses and any difference in headshaking frequency between the first and second test occasion still be valid as a measure of the respective stress levels. 

All horses learned to distinguish between the rewarded and unrewarded bucket locations quicker on the second test occasion, after the treatment period, indicating that they remembered the exercise eight to nine weeks after learning it. There was no significant difference between the two groups and therefore no treatment effect on how well the horses remembered the task.

In previous studies horses trained with positive reinforcement learned a task faster and remembered it better than horses trained with negative reinforcement. As the cognitive bias test is designed to teach horses to distinguish between locations by using food rewards it could be argued that the horses learn the task through positive reinforcement, and this would then support previous findings that tasks learned using positive reinforcement are remembered even after a long time.

The treatment horses did not approach the ambiguous bucket faster during the second test occasion, indicating that the positive reinforcement training period of eight weeks did not improve their optimism. Previous studies found that individuals that move from a negative to a positive environment, for example rescued horses, become more optimistic. It is possible that the treatment period was too short or not intensive enough for the treatment horses to perceive that their environment changed positively, as the owners were only asked to perform five minutes of positive reinforcement training four times a week.

Surprisingly, looking at the absolute latencies to approach the buckets before and after the treatment period the control horses were quicker to approach the ambiguous location on the second test occasion, although the difference in latency was not significant. A previous study found, similarly, that mares trained with negative reinforcement were more optimistic in a cognitive bias test than mares trained with positive reinforcement.

One possible explanation for this is that the value of the bucket might have been higher to the control horses who had not been given any positive reinforcement training during the treatment period and thus were more motivated to explore the buckets. It is also possible that they perceived the test situation as more positive in relation to their previous experiences, causing an optimistic judgement bias. 

While cognitive bias tests seem to be a valid method of assessing emotional state in animals there are some limitations to the design that may influence the results. A review by Lagisz et al. (2020) suggests that a test design that measures a simple go/no go response to an ambiguous stimulus is more sensitive to habituation effects and learned responses to cues in the test setup that may overshadow the emotional state. The choice of reinforcements during the training phase can also influence the animal’s perception of the rewarded and unrewarded location. Pairing reward and no reward, as in this study, may give a weaker effect than pairing reward with an aversive in the unrewarded location.  As I performed all tests in the horses’ respective home environments on two different occasions, it is also possible that the results were confounded by uncontrollable distractions in the environment such as other horses leaving or arriving at the yard, cars, changes in weather on the test days et cetera.

I found no significant difference in hair cortisol concentrations before and after treatment for either group. As mentioned before, one possible reason is that the treatment period wasn’t long or intensive enough to affect the concentrations. The addition of five minutes of positive reinforcement training four times a week may not be enough to offset the effect of daily stabling and management routines on long-term stress levels. A previous study has found that hair cortisol concentrations can differ significantly between stables and I did not control for stabling when analysing my data.

It is also possible that the second hair sample was taken too early and thus shows the hair cortisol concentrations about halfway through treatment rather than after the full treatment period, as hair grows at a rate of about one centimetre per month.

Both treatment and control horses had higher hair cortisol concentrations after the treatment period compared to before. This could be due to seasonal changes similar to the ones seen in dogs, as the first sample was taken in September and the second sample in December. Salivary cortisol levels in horses seem to have a peak in December as well.

Interestingly, owners of the treatment horses reported lower scores for the relationship questionnaire both before and after the treatment period than owners of control horses. As the horses were divided into treatment and control groups pseudorandomly before the results from the tests or questionnaires were known, the groups were not balanced for relationship scores.

Treatment horses showed a tendency to improve their relationship score after the treatment period while control horses did not. This, along with the increase in contact-seeking behaviours in the treatment group, supports the idea that positive reinforcement training has a positive effect on the horse-human relationship.