Abstract: A behavioral assessment of horses who were being used and not used in therapeutic riding programs was conducted to help determine useful methods of selecting horses for use in therapeutic riding programs. A total of 103 horses (76 horses from five therapeutic riding centers and 27 non-therapeutic riding horses from four sites) were used. Temperament survey for each horse were completed by three riding instructors at each therapeutic riding center or by the individual most knowledgeable about the horse at the other sites. Twenty personality traits from the survey were used to quantify temperament. Concentrations of plasma cortisol, norepinephrine and epinephrine were also measured in each horse. A reactivity test was then conducted which involved introducing three novel stimuli: a walking and vocalizing toy pig placed on a cardboard surface in front of the horse for 20 s; popping a balloon near the horse's flank area; and suddenly opening an umbrella and holding it open in front of the horse for 20 s. Reactions (expressions, vocalizations and movement) to each of the stimuli were scored and used to calculate an average reactivity score for each horse. The therapeutic riding instructors did not often agree on the temperament of their center's horses. The personality trait ratings made by the therapeutic riding instructors at each center were on average significantly correlated (P<0.01, r>0.52) for only 37.8% of the horses for any two instructors and 7.8% for three instructors. No significant correlations were found between temperament, reactivity, and the hormone concentrations (r<0.19), but regression analysis indicated a possibility of predicting temperament from the reactivity score and hormone concentrations (P<0.08). There was also a tendency for relationships between extremes in temperament (desirable vs. undesirable) and the hormone concentrations (P<0.09), and between extremes in reactivity (low vs. high) and the hormone concentrations (P=0.08). The difference in ratings among riding instructors indicates a need for more collaboration between instructors when evaluating horse temperament. This study also indicates that it was very difficult to objectively determine the suitability of horses for therapeutic riding programs regarding their temperament and reactivity, probably because other traits (e.g., smoothness of gait) are also considered very important.