Fusion of vestibular and podokinesthetic information during self-turning towards instructed targets

When observers step about their vertical axis ("active turning") without vision they dispose of essentially two sources of information that can tell them by how much they have turned: the vestibular cue which reflects head rotation in space and the "podokinesthetic" cue, a compound of leg proprioceptive afferents and efference copy signals which reflects the observer's motion relative to his support. We ask how these two cues are fused in the process leading to the perception of self-displacement during active turning. To this end we compared the performance of observers in three angular navigation tasks which differed with regard to the number and type of available motion cues: (1) Passive rotation, vestibular cue ( ves) only; observers are standing on a platform which is being rotated. (2) Treadmill stepping, podokinesthetic cue ( pod) only; observers step counter to the rotating platform so as to remain stable in space. (3) Active turning, ves and pod available; observers step around on the stationary platform. In all three tasks, angular velocity varied from trial to trial (15, 30, 60 degrees /s) but was constant during trials. Perception was probed by having the observers signal when they thought to have reached a previously instructed angular displacement, either in space or relative to the platform ("target"; range 60-1080 degrees ). Performance was quantified in terms of the targeting gain (displacement reached by the observer divided by target angle) and of the random error ( E(r)), which records an observer's deviation during single trials from his average performance. Confirming previous observations, E(r) was found to be significantly smaller during active turning than during passive turning, and we now complement these observations by showing that it is also significantly smaller than during treadmill stepping. This behaviour of E(r) is compatible with the idea that ves and pod be averaged during active turning. On the other hand, the observed characteristics of the targeting gain ( G(T)) support this idea only for the case of fast rotations (60 degrees /s); at lower velocities, the gain found during active turning was clearly not the average of the G(T) values recorded in the passive and the treadmill modes. We therefore also discuss alternative scenarios as to how ves and pod could interact, among these one based on the concept of a vestibular eigenmodel. A common denominator of these scenarios is that ves assumes the role of a prerequisite for an optimal use of pod during turning on a stationary support, without itself entering the calculation of displacement perception; this perception would be based exclusively on pod. Finally, it was a consistent observation that during passive rotations cognitive mechanisms fill in for the decaying vestibular signal in the context of the present navigation task, enabling observers to achieve large displacements surprisingly well although the duration of these movements exceeds by far the conventionally cited value of the central vestibular time constant (=20 s).