Twenty horses were paired for age, sex and breed and placed into one of two groups. The horses in Group A (control) were reinforced using only negative reinforcement (NR) while those in group B (treatment) were reinforced with both positive reinforcement (PR) and NR concurrently. All horses were shaped for the halt response while being driven in long-reins over a period of 5 consecutive days. On day I, all horses were given a baseline test of 20 random halts while being long-reined in an indoor arena. On days 2-4, the shaping of the halt response continued with horses being reinforced according to the group to which they had been allocated. On day 5 of testing the baseline test was repeated (final test). During the baseline and final tests, behavioural responses and accuracy of completion of the halt response were recorded. Heart rates were recorded continuously during testing. One-way analysis of variance in randomised blocks and analysis of covariance using baseline data as a covariate showed no effect on latency to halt. However, horses reinforced with both NR and PR shook their heads vertically less and were more likely to lick their lips than those reinforced with NR only. There was also a trend for an increase in roundness of outline of the horses that were reinforced with both PR and NR. These results suggest that the implementation of PR effectively into equitation training may improve the welfare of the horse.