Abstract: The efficacy of questionnaire-based personality assessment has been shown in a variety of animal and human personality studies. There has been a recent increase in questionnaire-based studies focussing on equine personality but with a lack of comparability to studies on other species. The aim of this study was to test the reliability of an assessment method originally developed for primates and demonstrate reliability using three criteria (1) assessments by independent observers must agree with one another, (2) these assessments must predict behaviours and real-world outcomes and (3) observer ratings must be shown to reflect genuine attributes of the individuals rated, not merely the observer's implicit personality theories about how traits co-vary. The personality of 61 horses (Equus caballus) was assessed using a questionnaire constructed of 30 behaviourally defined adjectives (BDAs). Horses were each assessed by three judges, in addition to a total of 2 h behaviour observations recorded per horse. Rater agreement was demonstrated for 72.1% of the horses and 25 of the BDAs. Principal component analysis was carried out on the rating data and revealed six underlying personality components that were labelled “dominance”, “anxiousness”, “excitability”, “protection”, “sociability” and “inquisitiveness”. Component scores for horses were correlated against behavioural observations for the same horses and revealed significant correlations with 20 of the recorded behaviour variables. Correlations between specific components and their associated behaviours were logical and consistent with the types of behaviours that would be expected to be linked with such personality types. The data were shown to meet all three criteria and provided strong evidence that the assessment method was reliably measuring horse personality.