Training results in marked modelling of the subchondral bone of the carpus, but the effect of training on the subchondral bone of the distal tarsal joints is unknown. The aim of this study was to determine whether training influenced modelling of the third and central tarsal bones in Thoroughbred horses. Twelve untrained Thoroughbred horses were divided into 2 groups. Group 1 underwent a 19 week progressive training regimen on a high speed treadmill. Group 2 were walked for 40 min daily. Images of left tarsi were obtained by magnetic resonance imaging (MRI) with a 0.5 Tesla superconducting magnet using a spin echo sequence. Sagittal and oblique sagittal slices were made perpendicular to the articular surfaces of the intertarsal joints and were analysed using image analysis software to measure the proportion of dense subchondral bone in the dorsal facet of each bone. Mean +/- s.d. percentage area of dense subchondral bone in the dorsal facet of the central tarsal bone in Group 1 was 47 +/- 8 medially, 46 +/- 4 sagitally and 50 +/- 11 dorsolaterally, whereas in Group 2 it was 39 +/- 16 medially, 43 +/- 8 sagitally and 53 +/- 7 dorsolaterally. For the third tarsal bone mean +/- s.d. percentage area of dense subchondral bone in Group 1 was 32 +/- 10 medially, 39 +/- 11 sagitally and 44 +/- 8 dorsolaterally, whereas in Group 2 it was 28 +/- 8 medially, 37 +/- 6 sagitally and 41 +/- 9 dorsolaterally. There was no significant difference in percentage area of dense subchondral bone between the trained and untrained horses. An effect of treadmill training of Thoroughbred horses on modelling of the central and third tarsal bones could not be demonstrated.