Variability of scores in the 2008 Olympic dressage competition and implications for horse training and welfare

Olympic dressage involves “an intimate unity between a human and a non-human” and is scored by a subjective judging process, under the auspices of the Fédération Equestre Internationale whose Code of Conduct declares the welfare of the horse as paramount. Dressage is of particular interest to equitation scientists and equine ethologists because it embodies the full range of the stimulus-response contingencies that operate in all of the Olympic disciplines. In Fédération Equestre Internationale dressage competition, collective marks are awarded across four domains immediately after each performance. Collective marks are designed to summarize the performance of horse and rider and must reflect the qualities of the entire performance. They are derived from the observation of the judges of the separate test movements. The 4 collective marks include: (1) paces; (2) impulsion; (3) submission; and (4) the rider\u27s position and seat; correctness and effect of the aids (rider signals). The definition of submission in this context makes reference to lightness and other qualities that align with optimal ridden horse welfare. We assessed the characteristics of these marks in horses competing in the 2008 Olympic Games Grand Prix (GP; n = 46) and Grand Prix Special (GPS; n = 25) dressage competitions. We also examined the effect of judge location and used Pearson correlation coefficients to explore relationships between collective marks and test-movement scores. All 4 collective marks correlated with each other significantly (P \u3c 0.001). The weakest correlation was between paces and submission (r = 0.22) and the strongest between impulsion and rider position scores (RPS) (r = 0.59). In the GP, paces and submission scores were less correlated with test movement scores than the impulsion and RPS scores. In the GPS, submission scores were less correlated with individual movements than the other collective marks. Indeed, they failed to significantly correlate with 19 of 32 movement scores (P \u3c 0.05). RPS varied most in the GP (standard deviation = 0.73) whereas submission scores varied most in the GPS (standard deviation = 0.65). A REML analysis across both competitions showed all collective marks were significant in predicting final percentage scores but submission (F = 31.27) made the least significant contribution (paces, F = 61.3; impulsion, F = 69.77; RPS F = 53.01; P \u3c 0.001 for all values). These results speak of considerable variability in judging and suggest that, despite the relevance of submission to horse welfare, judges have considerable difficulty scoring in this domain and aligning their scores with overall performance