Spatial weighting of orientation ensembles is shifted towards highly correlated regions

Poster Presentation: Sunday, May 18, 2025, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Scene Perception: Ensemble

Richard Bailey1, Jefferson Ortega1, Andrey Chetverikov2, David Whitney1; 1University of California, Berkeley, 2University of Bergen

Observers can easily report the average orientation of an array of oriented lines with a randomly generated mean orientation. Previous papers showed that one can calculate a spatially weighted map from these ensemble judgments, which reveals specific elements or regions in the display that matter most for ensemble orientation perception (Tiurina et al., 2024). In the experiments here, we used this approach to test how correlations in the visual environment affect ensemble perception. Observers judged the mean orientation of an ensemble of lines. On a given trial, the orientations were drawn from the same underlying distribution with a common mean, and the only manipulation was whether the orientations were more correlated in one half of the display (left or right). In Experiment 1, the correlated side of the display was held constant on one side of the screen on every trial. Our results showed that the spatially weighted maps shifted toward the correlated side. This hints that correlated information in the world might modulate ensemble perception. One possibility, though, is that when there is more correlation there is also lower variance in the display. So, observers might use variance as a proxy for correlation, and the spatially weighted maps may simply be biased toward lower variance in the display. To address this, in Experiment 2, the correlated side was randomized in either left or right visual field, from trial to trial, unpredictably. Interestingly, we found that the spatially weighted maps did not shift toward the lower variance (correlated) side. This suggests it is not a trial-by-trial shift in spatially weighted maps toward lower variance in the display. Instead, it seems to be a spatially weighted bias toward higher correlations in the environment, which builds up over time.

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