Minimising bone mass is a generally accepted requirement for efficient locomotion. However, the young Thoroughbred racehorse needs to develop sufficient skeletal strength to withstand the stresses of training and racing.  The modelling of the dorsal cortex of the MC3 bone is an essential part of the racehorse's adaptation to fast work. The effects of race preparation on the MC3 were monitored in 6 yearling Thoroughbred horses, trained for racing in an incremental exercise program over 5 months.  Six untrained yearlings were kept as controls. Intravital fluorescent bone markers that enabled bone growth on the midshaft dorsal surface of MC3 to be measured from bone biopsies taken at the end of the experiment were infused at 5 week intervals. They showed that 15 to 30 strides of exercise >12 m/s were associated with bone modelling in this area.  Only 30 strides faster than 12 ms-1 repeated four times in five weeks were sufficient to produce a maximal response (mean of 6 exercised horses 1.59 ± 0.9 µmeters per day during gallop exercise, compared with 0.97 ± 0.5µmeters per day in the 6 controls which were free to exercise themselves in the paddock).  The amount of modelling of the MC3 could be related to differences between the horses in the shape of the bone in the dorsopalmar plane, and not to measurements of bone quality such as bone mineral density and ultrasound speed.  Only one horse went shin sore in the experiment.  This horse was subjected to 60 strides of fast exercise and went shin sore during a second session two days later. This horse showed a disrupted arrangement of new bone on the dorsal surface, compared to the plexiform bone formed by the horses showing the maximal bone response.