Inverse dynamics analyses are commonly used to understand movement patterns in all forms of gait. The aim of this study was to determine the effect of digital filtering procedures on the knee joint moments calculated during sprinting as an example of the possible influence of data analysis processes on interpretation of movement patterns. Data were obtained from three highly trained sprinters who completed a series of 30 m sprints. Ten different combinations of cut-off frequency were applied to the two-dimensional kinematic and kinetic input data with the kinetic cut-off frequency set equal to or higher than the kinematic cut-off frequency. When using the commonly adopted practice of filtering the kinetic data with a higher cut-off frequency than the kinematic data, exaggerated fluctuations in the knee joint moment existed soon after contact. In extreme cases, the knee moved between flexor, extensor and flexor dominance in less than 33 ms and through ranges exceeding 500 Nm. During an inverse dynamics analysis of locomotion, mismatched cut-off frequencies will likely affect the calculated joint moments if the cut-off frequency applied to the kinematic data is less than the true frequency content, particularly during impact phases. In the example of sprinting, exaggerated fluctuations in the knee joint moment appear to be data processing artefact rather than genuine characteristics of the joint kinetics. When the cut-off frequencies, and thus the frequency content of all input data, are matched, the fluctuations after contact are minimal and such a procedure is suggested for inverse dynamics analyses of gait.