Abstract: The benefits to humans of equine-assisted therapy (EAT) have been well-researched, however few studies have analyzed the effects on the horse. Understanding how differing mental states of humans affect the behaviour and response of the horse can assist in providing optimal outcomes for both horse and human. Four humans clinically diagnosed and under care of a psychotherapist for Post-Traumatic Stress Disorder (PTSD) were matched physically to four neurotypical control humans and individually subjected to each of 17 therapy horses loose in a round pen. A professional acting coach instructed the control humans in replicating the physical movements of their paired PTSD individual. Both horses and humans were equipped with a heart rate (HR) monitor recording HR every 5secs. Saliva samples were collected from each horse 30 min before and 30 min after each trial to analyze cortisol concentrations. Each trial consisted of 5 min of baseline observation of the horse alone in the round pen after which the human entered the round pen for 2 min, followed by an additional 5 min of the horse alone. Behavioural observations indicative of stress in the horse (gait, head height, ear orientation, body orientation, distance from the human, latency of approach to the human, vocalizations, and chewing) were retrospectively collected from video recordings of each trial and analyzed using a repeated measures GLIMMIX with Tukey's multiple comparisons for differences between treatments and time periods. Horses moved slower (p < 0.0001), carried their head lower (p < 0.0001), vocalized less (p < 0.0001), and chewed less (p < 0.0001) when any human was present with them in the round pen. Horse HR increased in the presence of the PTSD humans, even after the PTSD human left the pen (p < 0.0001). Since two of the PTSD/control human pairs were experienced with horses and two were not, a post-hoc analysis showed that horses approached quicker (p < 0.016) and stood closer (p < 0.0082) to humans who were experienced with horses. Horse HR was lower when with inexperienced humans (p < 0.0001) whereas inexperienced human HR was higher (p < 0.0001). Horse salivary cortisol did not differ between exposure to PTSD and control humans (p > 0.32). Overall, behavioural and physiological responses of horses to humans are more pronounced based on human experience with horses than whether the human is diagnosed with a mental disorder. This may be a reflection of a directness of movement associated with humans who are experienced with horses that makes the horse more attentive. It appears that horses respond more to physical cues from the human rather than emotional cues. This knowledge is important in tailoring therapy programs and justifying horse responses when interacting with a patient in a therapy setting.