Abstract
Human perception is imperfect. For instance, subjects’ perception of a target’s location in the current trial would be systematically biased toward that in previous trial, i.e., ‘serial dependence effect (SDE)’ in spatial perception. Meanwhile, instead of presenting stimulus in a continuous two- or three-dimensional space as we typically experience in daily life, previous studies have limited the stimulus presentation on a certain circle, leaving the SDE in a continuous spatial context largely unknown. Here, we performed three experiments to examine the SDE in two-dimensional, continuous space and the underlying spatial organization principle. In each trial, participants were presented with a stimulus at specific location and needed to reproduce its location later for all three experiments varying in spatial context (distribution within round area or square area), task difficulty (with or without an irrelevant task), stimulus shape (dot or bar), and experimental environment (online or offline). We used Cartesian (x and y) and Polar (ρ and ϕ) coordinates to characterize SDE: the response error in a certain dimension as a function of distance between the stimuli locations in current trials and those in previous trials. First, for the SDE in ϕ dimension, we found that the SDE was best explained by the DoG (derivative of gaussian) function, similar to previous findings for other circular variables, e.g., location on a circle, orientation, etc. In contrast, the SDE in other dimensions (x, y, ρ) showed a linear function. After combing two individual dimensions into coordinates, we found Cartesian coordinates outperformed Polar coordinates in explaining the SDE consistently among all three experiments. Taken together, by assessing the SDE of spatial location in a continuous space, we find that Cartesian coordinates (horizontal and vertical axes) instead of Polar coordinates (diagonal axes) are employed to organize continuous spatial information over time.