Abstract: Four methods of ranking horses on reactivity were evaluated and compared: isolation from conspecifics, presentation of a static novel stimulus, traversing a novel stimulus in a runway (isolation, novel stimulus and runways tests, respectively) and assigning subjective emotionality scores. Forty horses performed each of the three tests daily on three different days in a switchback design where treatments were injection of a tranquilizer or vehicle. Horses were randomly assigned a daily test sequence, which was maintained throughout the study. In all tests, heart rates were recorded and behavior was videotaped. To be considered a valid test of reactivity, at least one heart rate and one behavioural measurement in the test had to show a significant difference due to tranquilization, and behavioural measures had to be displayed in at least 75% of the trials. In the runway test, no significant difference in heart rate values in tranquilized and non-tranquilized horses was found, and no behavioural attribute was displayed in more than 52% of the trials; therefore it was rejected as a valid test of reactivity. Both isolation and novel stimulus tests produced valid measurements. Mean heart rate was the most precise physiological measure for these tests, and walking and defecation frequency were the most precise behavioural measures for novel stimulus and isolation tests, respectively. Mean heart rates on the novel stimulus and isolation tests were correlated (rs=0.79, P<0.01) indicating that these tests produced similar rankings based on physiological responses. However, behavioural measures ranked horses differently (rs=0.27, P<0.10) on the tests. Rank correlations between mean heart rates and behavioural measures were higher in the novel stimulus (rs = 0.66, P<0.01) than the isolation test (rs = 0.55, P<0.01), indicating that the novel stimulus test ranked horses based on either physiological or behavioural responses more similarly than did the isolation test. Therefore, the novel stimulus test was considered the more accurate evaluation of reactivity. Subjective emotionality scores were correlated moderately with mean heart rates (rs > 0.33, P<0.01) from the novel stimulus and isolation tests and with walking scores (rs = 0.47, P<0.01) from the novel stimulus test. Assignment of subjective emotionality scores was not as accurate as the novel stimulus or isolation tests in ranking horses for reactivity. Using physiological data alone, combining physiological and behavioural measurements or using more than one behavioural measurement in reactivity tests may reflect the reactivity of the horse better than a single behavioural measurement.