Visual control of trunk translation and orientation during locomotion

Abstract

Previous studies have suggested distinct control of gait characteristics in the anterior-posterior (AP) and medial-lateral (ML) directions in response to visual input. Responses were larger to a ML visual stimulus, suggesting that vision plays a larger role in stabilizing gait in the ML direction. Here, we investigated responses of the trunk during locomotion to determine whether a similar direction dependence is observed. We hypothesized that translation of the trunk would show a similar ML dependence on vision, but that angular deviations of the trunk would show equivalent responses in all directions. Subjects stood or walked on a treadmill at 5 km/h while viewing a virtual wall of white triangles that moved in either the AP or ML direction according to a broadband input stimulus. Frequency response functions between the visual scene motion and trunk kinematics revealed that trunk translation gain was larger across all frequencies during walking compared with standing. Trunk orientation responses were not different from standing at very low frequencies; however, at high frequencies, trunk orientation gain was much higher during walking. Larger gains in response to ML visual scene motion were found for all trunk movements. Higher gains in the ML direction while walking suggest that visual feedback may contribute more to the stability of trunk movements in the ML direction. Vision modified trunk movement behavior on both a slow (translation) and fast (orientation) time scale suggesting a priority for minimizing angular deviations of the trunk. Overall, trunk responses to visual input were consistent with the theme that control of locomotion requires higher-level sensory input to maintain stability in the ML direction.

Fig. 1

Illustration of the experimental setup. Subjects stood or walked on the treadmill in front of a virtual display of randomly oriented white triangles on a black background. Subjects wore goggles which prevented them from seeing the borders of the TV

Fig. 2

Exemplar time and frequency content of visual scene motion shown for 60 s from a single trial. The position of the visual scene in the static condition corresponded to the zero value during trials with visual scene motion. Positive (negative) values indicate that the vir tual wall of triangles was “moving away” (toward) from the subject during the AP stimulus conditions. During ML stimulus conditions, positive (negative) values indicate movement of the virtual wall of triangles to the right (left)

Fig. 3

Gain and phase plots for condition comparison, only frequency bins significantly different from zero have been plotted. Frequency response functions (gain and phase) for translation: a lumbar gain, b lumbar phase, c neck gain, d neck phase, and orientation: e trunk gain, f trunk phase in response to visual scene motion. Error bars represent bootstrapped standard error

Fig. 4

Coherent and incoherent position (subplots a and b) variance. Open squares represent variance incoherent with dynamic visual scene motion and filled squares represent variance coherent with dynamic visual scene motion. Circles represent total variance with stationary visual scene. AP and ML references on the x-axis correspond to both kinematic direction and visual stimuli direction. Comparisons between incoherent variance for dynamic visual scene conditions and total variance for the static visual scene condition are shown in subplot b. Lumbar refers to translation, while trunk refers to trunk angle. Error bars represent standard error of the mean