A statistical approach to sensorimotor strategies: conjugate cross-correlations

A simple method, based on cross-correlation functions (CCFs) between two time series of kinematic or physiological measurements, is proposed for the analysis of multisegmental movements. Special emphasis is placed on measuring accelerations. When the movements of two body segments are coordinated but consistently time lagged, their CCF displays a peak at the corresponding time abscissa. The reproducible positions of the peaks reflect biomechanical or physiological constraints. Several significantly large peaks can be observed in a CCF. It is possible to identify coordinated movements involving more than two segments by applying simple rules of compatibility between the time lags and between the signs of the correlation peaks. With the method proposed, it is possible to determine the signs of relative variation and the time lags of the successive statistically correlated segmental movements. This is particularly useful in the case of both continuous and periodic sensorimotor control, where classical poststimulus methods cannot be applied. Unlike the classical poststimulus methods, this method does not require a time origin, and it is not necessary to monitor the muscles or even to specify exactly which ones are involved. The method is also applicable to experiments involving a time origin (e.g., and applied perturbation), although in this case it is less accurate than the averaging technique. Individual postural strategies can be identified, which suggests some interesting potential applications of the method to clinical studies.