A gait smoothness index has been used to identify differences in walking gait between healthy and disease populations, but has not yet been applied to running gait, where it may indicate skill differences. The index quantifies the relative level of out-of-phase harmonics in the frequency spectrum. PURPOSE: To use a cross-sectional study design to discriminate running smoothness in four groups of healthy, young subjects. METHODS: We measured 3-D trunk acceleration using optical motion capture (Vicon, Centennial, CO) at 200 Hz for 60-90 s at 3.33, 3.89, and 4.44 m·s-1. Subjects included: 1) NCAA Division I cross-country runners (n=11); 2) high school cross-country runners (n=19); 3) untrained subjects with BMI < 25 (n = 8); and 4) untrained subjects with BMI > 25 (n=8). We performed a fast Fourier transform of the mediolateral (ML), anterioposterior (AP), and vertical (VT) acceleration signal and calculated the amplitude of the first 20 harmonics. AP and VT are dominated by the second harmonic, so odd numbered harmonics (out-of-phase accelerations) detract from smoothness. The smoothness index is the ratio of the sum of the even harmonic amplitudes to the sum of the odd harmonic amplitudes. Since the first harmonic dominates ML, the smoothness index for this axis is the ratio of odd to even. A higher ratio indicates greater smoothness. We used a group×speed ANOVA to test for main effects, with a Bonferroni post hoc test where appropriate. Analyses were performed in Matlab (2013b, Mathworks, Natick, MA). RESULTS: College and high school runners had higher AP smoothness than untrained-low BMI and untrained-high BMI (6.74 and 7.54 versus 4.33 and 4.43, respectively, p<0.05). For VT, college and high school runners had higher smoothness than untrained-high BMI (10.91 and 10.26 versus 6.75, respectively, p<0.05). High school runners also had a higher VT smoothness than untrained-low BMI (10.26 versus 8.32, respectively, p<0.05). There were no speed effects or any group×speed interactions (p>0.05). CONCLUSION: Running smoothness was significantly different between trained and untrained subject groups. It is unknown if this is due to training effects on biomechanics and motor control, training-associated body composition changes, or pre-existing characteristics leading to self-selection of running, all of which might change the contribution of out-of-phase movement to the AP and VT acceleration signal.