Theo Schmit¹, Jefferson Ortega¹, Andrey Chetverikov², David Whitney¹; ¹University of California, Berkeley, ²University of Bergen

Serial dependence stabilizes our perceptual experience by biasing our current perception toward previously seen stimuli. While this effect is well documented for simple features, less is known about how variance in ensemble presentations influences serial dependence. We investigated this question by manipulating stimuli variance in an ensemble orientation judgment task. Participants were presented with 36 oriented lines and tasked with reporting the overall, average orientation of all lines. There were two types of trials in our experiment. In asymmetric variance trials, 16 orientations on the left side of the screen were more correlated to each other than the 16 orientations on the right side of the screen, leading to an asymmetry in orientation variance across the screen. In symmetric variance trials, stimuli variance was uniform across the entire screen. Participants completed 800 trials (400 for each trial type) and trial type was interleaved throughout the experiment such that asymmetric variance trials always followed symmetric variance trials and vice-versa. For asymmetric variance trials, we found that observers made larger overall errors and they exhibited a significant serial dependence effect, suggesting that ensemble judgments were biased towards their perception of the previous trial orientation. We found significantly less serial dependence and more accurate reports for symmetric variance trials. The results suggest that when performance levels are low, potentially indicating higher uncertainty of the estimated ensemble, observers rely more heavily on past experiences to inform their current judgments.

Acknowledgements: Supported in part by the National Institute of Health (grant no. R01CA236793) to D.W and  (grant no. 1F99NS141343) to J.O.