C. Stella Qian¹ (), Jian-Qiao Zhu², Nick Chater³, Adam Sanborn¹; ¹Department of Psychology, University of Warwick, ²Department of Computer Science, Princeton University, ³Warwick Business School, University of Warwick

Perceptual decision making does not terminate when the stimuli are off. Instead, previous work has found that a discrimination task could bias the later perceptual judgement task away from the reference in former task, which is termed the repulsion effect (Zamboni et al., 2016). The repulsion effect has been separately explained as an effect of self-consistent inference (Luu & Stocker, 2018), or as the result of optionally stopping an internal sampling process when the discrimination is the clearest (Zhu et al., 2023). Here, we explored the impact of more time between the two tasks on the strength of the repulsion effect, which has consequences for both theoretical positions. Participants were presented with a random dot motion stimuli and asked to do two tasks sequentially: a motion direction discrimination task relative to a probe stimulus presented after the motion stimuli and then a motion direction report task with the mouse cursor. We found better accuracy and faster reaction time as the difference between motion direction and probe stimulus became larger. The motion direction report task also had good accuracy and precision (absolute deviation M = 13.77 °, SD = 36.23°). The repulsion effect was replicated: response deviation conformed to a bimodal distribution when the discrimination judgement task was at chance level; the reported direction relative to the probe also conformed to a bimodal distribution. We tested the effect of adding a pause (1.5s and 2s) after the discrimination task results while keeping the probe onscreen. The repulsion effect decreased with a pause but did not disappear. The reference repulsion effect decreased with a delay between the discrimination task and the judgement task. We speculate that either self-consistency becomes less important with a delay, or the internal sampling process of the motion signal continues after the discrimination is made.

Acknowledgements: This work is supported by a European Research Council consolidator grant (No. 817492-SAMPLING).