Implicit and dynamic tracking of temporal probability

Research has shown that repeated exposure to stable patterns or statistics allows the visual system to learn and exploit the regularities for behavioural optimisation. However, the real-world environment is dynamic, and underlying statistics can frequently shift over time. Here, we show that the visual system can indeed track such dynamic probabilistic structure independent of our conscious awareness. Across five experiments utilising a spatial cueing paradigm, we systematically manipulated cue validity in a fine-grained manner not disclosed to participants. We also measured and even manipulated subjective awareness of the cue-target association using binary report, priming and self-estimation paradigms. In Experiments 1-3, the reaction time for the invalid cues increased with higher cue validity, suggesting that the visual system dynamically track cue validity by penalising invalid cues more heavily. Crucially, post-hoc binary reports revealed that the subjective awareness of the cue-target association had little impact on the reaction time. In two follow-up experiments, we used an explicit prime and asked participants to estimate the cue validity directly. The results across the different tasks consistently showed that reaction time reflected the actual validity, not the perceived/primed validity, further indicating that the tracking could be independent of conscious estimation. Together, we conclude that the visual system can implicitly and dynamically track detailed changes in temporal statistics that cannot be accessed by our conscious reasoning.