Abstract
An outstanding question regarding serial dependence is the level at which it operates—some argue for its direct influence on perception, while others posit that serial dependence only impacts post-perceptual processes. Here we examined when serial dependence appears and disappears to address this issue. We developed an ocular tracking task consisting of distinct temporal phases of tracking (retinal-motion-driven pursuit initiation vs. extraretinal-signal-supported steady-state tracking), coupled with high-resolution eye movement recordings, to provide an ideal testing paradigm for this purpose. Participants (N=16) tracked the step-ramp motion of a target spot (diameter: 0.64º; speed: 16º/s; direction: randomly drawn from the full 360° circle at the step size of 12°). We performed model-dependent analyses to measure the extent to which pursuit directions through the course of the current trial were affected by the previous target moving direction, with positive values indicating attraction and serial dependence and negative values indicating repulsion and adaptation. We observed a strong serial dependence at pursuit initiation that quickly declined over time, followed by a low-amplitude adaptation that remained stable throughout steady-state tracking. This result shows that serial dependence happens before adaptation in ocular tracking, providing evidence that serial dependence influences perception. We also found a strong correlation (r=0.88) between the strength of serial dependence and pursuit direction noise. Using a Bayesian observer model constrained by efficient coding, we further found that the temporal dynamics of serial dependence can be predicted by pursuit direction noise over time. This supports our proposal that the visual system may strike a balance between utilizing the temporal continuity of the physical environment (through serial dependence) and optimizing sensitivity to subtle changes in the environment (through adaptation), and this balance is regulated by sensory noise.