Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Rising Stars | Poster Sessions | Data Blitz

Investigating Leakage in Visual Perceptual Decision-Making

Poster Session A - Saturday, March 7, 2026, 3:00 – 5:00 pm PST, Fairview/Kitsilano Ballrooms

Eusabeia Silfanus¹, Aaron Schurger¹; ¹Chapman University

Previous research demonstrates that perceptual decision-making involves integrating noisy signals over time to a threshold. Such accumulator models may include leakage - the tendency for the decision variable to decay to zero in the absence of evidence. Although behavioral and brain data during perceptual decision-making can be well accounted for without leakage, this might be linked to common choices of stimulus and task. We investigate a possible role for leakage in visual perception, introducing a novel stimulus and task context where the behavior of a human observer will vary in a predictable way if the mechanism is leaky. Each frame of our dynamic noise stimulus consists of an oriented Gabor patch rendered using error-diffusion dither, with each pixel being either black or white depending, probabilistically, on the intensity of the corresponding pixel in the original Gabor. This allows us to precisely control the rate of information delivery without changing the physical intensity. The stimulus sequence consists of 20 unique noisy renderings of the same Gabor, presented at a certain rate and at a threshold-level information intensity. The task is to identify the orientation of the Gabor from among two options: leaning left or leaning right. If the visual system is inherently leaky then a percept will form only when the rate of evidence delivery overcomes the leakage, allowing the decision variable to reach threshold. Trial-by-trial accuracy and confidence judgments both decreased significantly as the evidence update interval increased. These results demonstrate a role for leakage in visual perceptual decision making.

Topic Area: PERCEPTION & ACTION: Vision