Multivariate statistical analysis of stride-timing measurements of nonfatigued racing Thoroughbreds

SUMMARY AND CONCLUSIONS Standard univariate and bivariate statistical methods and two multivariate methods, stepwise regression and factor analyses, were used to analyze 17 stride-timing measurements of 22 racing Thoroughbreds filmed at the beginning of races. All horses were in a right-lead transverse gallop. Data were extracted from frame-by-frame analysis of 16-mm film taken with a high-speed camera. The mean stride duration was 0.405 s and the mean stride frequency was 2.47 strides/s. Stride duration had a higher correlation to the suspension-phase duration (r = 0.864) than to stride-stance-phase duration (r = 0.644). The 2 components of stride, the suspension and stride-stance phases, lasted on average, 0.114 s and 0.291 s or 28.07% and 72.01% of the stride duration, respectively. The correlation between stride-stance and suspension phases was low (r = 0.1767). Timing of the forelimbs, rather than overall stance-phase duration, may influence the duration of the suspension phase because four of seven variables selected in a stepwise-regression analysis were exclusively from the forelimbs. The stance-phase duration, and therefore swing-phase duration, of all four limbs differed significantly (P < 0.05). The left forelimb (lf) and right hind limb (rh) had significantly shorter stance and single-support phases. Fore-limb and hind limb pairs differed in their timing and coordination. There were differences in advanced placement and overlap of the hind limbs, compared with that in the forelimbs. Significant (P < 0.05) differences were found between three limb overlap measurements. The overlap between the forelimbs was shortest and that between the hind limbs was the longest in duration. Overlap of the stance phases of the hind limbs was significantly (P < 0.05) longer in duration than that between forelimbs or that between the rh and lf. Stepwise-regression analysis of overlap between the stance phase of the rh and lf vs all other stride-timing measurements, except stride-stance- and suspension-phase durations, and stride duration selected two variables (advanced placement between forelimbs, single support of the lf stance phase) and gave a value of 0.9957 for the coefficient of determination (R2). Seemingly, this overlap was dependent on the timing of the lf. Total overlap lasted 0.076 s and comprised 18.81% of stride duration. The R2 value of total overlap vs suspension for real-time (s) data was 0.1632 and 0.0146 for percentage of stride data. These low values indicated that total overlap and the duration of the suspension phase were not related. Four factors were selected in factor analysis. These factors explained 86.56% of total variance. Factor 1, which explained 28.29% of total variance, represented the coordination between the lead hind limb and nonlead fore-limb. Factor 2 explained 24.32% of total variance and was interpreted to represent the synchronous action of contralateral limb pairs, particularly the forelimbs, and the dissociation between forelimb and hind limb timing. Factor 3 explained 17.96% of total variance and represented the coordination between timing of the hind limbs. Factor 4 explained 15.99% of total variance and represented the overlap between contralateral limb pairs. The key elements of the racing stride of Thoroughbreds were the coordination between the lead hind limb and nonlead forelimb and the synchronous action of contralateral limb pairs. From this analysis, it was predicted that contralateral limbs should act more synchronously and that the overlap between the lead hind limb and nonlead forelimb should decrease as velocity increases. Seemingly, at faster velocities, galloping Thoroughbreds will tend to rabbit-hop by using their hind limbs more in synchrony and by dissociating the action of these limbs from that of the forelimbs. This would result in more reliance on back flexibility.