Natalya Phillips¹, Luis D. Ramirez¹, John T. Serences¹; ¹University of California San Diego

Neural processing capacity is finite, necessitating efficient and predictive mechanisms for rapidly processing sensory information. One such mechanism is serial dependence, a phenomenon where perception is biased toward recent stimuli, especially under uncertainty. Here we investigated whether uncertainty, induced by altering contrast or orientation precision, influences the magnitude and tuning of serial dependence. Because contrast selectively alters neural response strength and orientation precision selectivity alters the dispersion of neural responses, we hypothesized that contrast primarily modulates the strength of serial dependence, and orientation precision primarily modulates tuning (width). We manipulated contrast or orientation precision within blocks of a 2AFC orientation discrimination task (orientation filter widths: 2°, 40°, 80°; contrasts: 90%, 50%, 25%). On every trial, an oriented bandpass-filtered noise stimulus (uniformly-sampled filter width or contrast level) was followed by a line that had an independently and uniformly-sampled orientation offset (7 offsets, linearly-spaced from 0–15°). Participants (n=11, each subject tested for 3–5, 1.5 hour sessions) reported whether the line was clockwise or counterclockwise from the filtered stimulus (2160–3510 trials per subject). Response bias was estimated with a Gaussian cumulative density function within bins of stimulus history (difference between the previous and current trial’s stimulus orientation; 32° sliding window) and for each condition (9 combinations of contrast and orientation precision levels). To quantify serial dependence, we estimated the amplitude and width of bias with a derivative of Gaussian function for every combination of uncertainty. We found that the magnitude of attraction decreased with certainty (though less so for contrast). Moreover, the width of attraction decreased with orientation precision, but increased with contrast. Our findings suggest that serial dependence is primarily influenced by current sensory evidence and less so by the strength of evidence on previous trials, with mechanistic differences under different types of uncertainty.

Acknowledgements: This research was funded by National Institutes of Health Grant R01EY025872 to J.T. Serences and supported by K00EY036804 to L.D. Ramirez and R25NS119707 to N.K. Phillips