Abstract: Highly-trained dressage horses were studied to test the hypothesis that stride length is altered independently of stride duration in the transitions between the collected, working, medium and extended trot. Six well-trained dressage horses were filmed at a frame rate of 150 frames/s performing the collected, working, medium and extended trots in a sand arena. Temporal, linear and angular data were extracted from the films, with 4 strides being analysed for each horse and gait type. There were no significant asymmetries between the left and rights limbs or diagonals when data from the whole group were pooled, but 3 horses showed asymmetries in one or more variables (P < 0.01). Analysis of variance and post-hoc tests indicated that the speed increased significantly (P < 0.01) from the collected (3.20 m/s) to the working (3.61 m/s) to the medium (4.47 m/s) to the extended (4.93 m/s) trot. The increases in speed were associated with a significant increase in stride length from 250 cm in the collected trot, to 273 cm in the working trot, 326 cm in the medium trot and 355 cm in the extended trot (P < 0.01). The lengthening of the stride was a result of increases between each gait type in the over-reach distance, whereas the diagonal distance was significantly longer in the extended than the collected trot only (P < 0.01). The stride duration tended to decrease as speed increased, and the difference became significant between the collected and extended trots (P < 0.01).