Abstract
Kalman filter models have been proposed to analyze integration of information in manual tracking of human beings. The idea of optimal control behind Kalman filter models, however, insights us to further investigate whether the predictability of object motion trajectories, as a kind of prior knowledge, can modulate information integration during continuous manual tracking. Specifically, we hypothesize that the Kalman gain, which refers to the relative contribution of visual input at the current time step in information integration, would decrease when motion trajectories become more predictable. In the current study, we generated three kinds of 2D trajectories, namely Gaussian noise trajectories, phase-free sinusoidal-wave trajectories and phase-locked sinusoidal-wave trajectories, respectively. For Gaussian noise trajectories, the displacements of target position at each time step followed a Gaussian distribution, N (0, σ2), where σ equaled to 2px. So, in this condition, trajectories were theoretically unpredictable. For two other kinds of trajectories, they were both additions of 7 nonharmonic sinusoidal waves with different periods (0.058s - 5s). The amplitudes of the sinusoidal waves were carefully adjusted so as to match three kinds of trajectories in displacement at each time point. 36 subjects participated the experiment and the tracking performances on each trial were fitted by a Kalman filter model separately. We used a Gibbs sampling method to estimate the posterior distributions of the Kalman gain in different conditions and found that the estimated Kalman gain decreased by 2% for both kinds of sinusoidal-wave trajectories than for Gaussian noise trajectories, where the estimated perceptual noise for inputs were comparable in three conditions. This finding demonstrates that people can modulates their information integration during manual tracking even without input noise change and suggests that an optimal control of manual tracking should not only take visual inputs during tracking but knowledge prior to them into account as well.