Daniel Yu¹ (), Jovan Kemp², Fulvio Domini¹; ¹Brown University, ²New York University Abu Dhabi

We depend on our visual speed perception to guide actions in dynamic environments and interpret how physical events unfold. Although we might assume that we can accurately perceive the speeds of moving objects, our speed estimates are often biased by other stimulus properties such as contrast: notably, low-contrast stimuli appear slower than high-contrast ones, especially at lower simulated speeds. Under probabilistic models of speed perception, low-contrast stimuli produce noisier estimates that yield biased responses when combined with a slow-speed prior. However, it is plausible that low-contrast stimuli may directly induce smaller estimates without being correlated with increased noise. To test this hypothesis, we conducted an experiment measuring speed discriminability for gratings with different contrasts. Two gratings were presented sequentially in each trial, and the participants judged which appeared faster. For a given run, the speed of one stimulus was held constant while the other was adjusted by an adaptive staircase procedure; we refer to the fixed-speed stimulus as the reference and the adjusted-speed stimulus as the comparison. If contrast does modulate estimation noise, then we expect that the discrimination thresholds for stimuli pairs where one is high-contrast and the other is low-contrast should not significantly differ. However, we found that the thresholds depended on the contrast of the comparison, regardless of the reference: high-contrast comparisons yielded smaller thresholds, while low-contrast comparisons yielded larger ones. These results suggest that estimation biases are directly related to visual stimulus quality, given by grating contrast, which may modulate the slope of the response function rather than the noise of visual estimates.

Acknowledgements: This research is supported by the National Science Foundation under Grant #2522066.