Zoe Little¹, Colin Clifford²; ¹University of New South Wales

Assimilative serial dependence in orientation perception occurs where the response to an orientated stimulus is biased towards the orientation of stimuli seen in the recent past. It remains a matter of debate whether serial dependence is mediated at the sensory or decisional level. Manipulations that induce uncertainty in decision-making without affecting the quality of sensory information may help to distinguish between these explanations. Here, we made use of the positive evidence bias, wherein increasing the positive evidence of a stimulus (i.e., overall contrast) while keeping the signal-to-noise ratio (SNR) the same increases the confidence but not accuracy of responses. Previous research has demonstrated that serial dependence in orientation perception is greater to high positive evidence stimuli, suggesting that perception may be skewed towards more certain previous decisions (Samaha et al. [2019] J. Vis.). We asked how decisional uncertainty (positive evidence) and sensory uncertainty (SNR) in both current and prior stimuli modulate the size of the serial dependence effect. Participants (N = 33) viewed oriented Gabor patches presented under different levels of SNR and positive evidence and reported their orientation, as well as their response confidence. We found that serial dependence was greatest when the stimulus on the target trial had low SNR, low positive evidence, or was responded to with low subjective confidence, affirming that uncertain stimuli are more likely to be integrated with previously seen information. We found no modulation of serial dependence by any source of uncertainty on the inducer trial. This has implications for theoretical accounts of serial dependence as well as for identifying whether serial dependence is driven by previous visual inputs or previous decisions.