Visual motion prediction under uncertainty must rely on both statistical and kinematic properties of the stimulus. Here, we investigated how decision-making processes and psychophysical parameters are modulated during extrapolation of random trajectories with different noise characteristics (Random Walk, RDW, or Independently and Identically Distributed, IID). Noise was applied to the horizontal position of a dot moving downward with constant vertical speed and vanishing before reaching the edge of the screen. Participants had to judge whether the dot would reach the edge right or left of the center. In Experiment 1 we varied the side of the last visible horizontal position, optimal for RDW extrapolation, and the mean of all visible positions, optimal for IID, to be either on the same or on opposite sides of the screen center. Experiment 2 investigated how the final segment of an IID path impacts the trajectory extrapolation when the last visible position and the mean of the last segment are on opposite sides of the center. Experiment 3 focused on assessing the accuracy of trajectory perception amid varying levels of noise. Behavioral and DDM (Diffusion Decision Model) analyses revealed that for RDW trajectories, participants relied on the last visible position, reflecting the temporal continuity of the path and leading to faster and more accurate decision making. IID trajectories showed greater variability in prediction strategies, with participants also focusing more on the last segment, as with RDW, rather than the mean position of the whole previous trajectory. However, this strategy works well even for IID paths despite being a suboptimal solution. These findings suggest that the perceptual system favors smooth motion for visual interpretation, aiding in the prediction of uncertain visual trajectories.
