We examined 106 Thoroughbred (TB) and 51 Standardbred (SB) racehorses that were presented because of poor performance. All horses were in training and had raced within 2 weeks of presentation. Horses were acclimatised (2–4 runs) and tested on a treadmill set at a 10% slope. The exercise test consisted of 3 min at 4 m/s, 2 min at 6 m/s, then 1 min steps at 8, 10, 11, 12 and 13 m/s or until the horse could not maintain pace with the treadmill. At the end of each speed, arterial blood and expired gas samples were collected and heart rate recorded. Measurements during the test included: arterial blood gas tensions, plasma lactate concentration [LA], V̇O2, V̇CO2, O2 pulse (V̇O2/HR), heart rate, maximum stride length (SLmax), post exercise haematocrit (PCVPE), total red cell volume (CV) and total run time. The TB ran longer than SB (522 vs. 477 s) and had higher values for CV, PCVPE, V̇O2max, V̇O2-200, maximum O2 pulse, V200, VHRmax, post exercise [LA] ([LAPE]) and SLmax. There were no breed differences for VLA4 and PaO2 at 10 m/s. In the TB, significant correlations (Spearman's rank order) (P<0.01) were found for treadmill run time with: PCVPE, V̇O2max, V̇O2–200, peak O2 pulse, PaO2 at 10 m/s, PaCO2 at 10 m/s, V200, VLA4, [LAPE], VHRmax, maximum blood temperature and SLmax. In the SB significant but lower correlations (P<0.05) were found for treadmill run time with: PCVPE, V̇O2max, V200, VLA4, [LAPE], VHRmax, maximum blood temperature and SLmax. While these results may not be applicable to normal racehorses, the indices which correlated best with treadmill run time were generally those indicating a high capacity for oxygen transport.