Testing the metric, topological, and visual hypotheses. A) High density: Soft metric neighborhood (dotted boundary) predicts decreasing influence of neighbors (shading) with metric distance from a pedestrian (bottom), whereas topological neighborhood (dashed lines) predicts decreasing influence with a neighbor's ordinal rank. Metric and topological distances are correlated here. B) Low density: The hypotheses are dissociated by manipulating crowd density. The metric neighborhood predicts that increasing distance will weaken neighbor influence, whereas the topological neighborhood predicts their influence will remain constant. C) In a visual neighborhood, influence decreases with both metric distance and visual occlusion. Contrary to the topological model, influence depends on density; contrary to the metric hypothesis, the model generalizes to crowds with different densities and configurations. Modified from Dachner, et al., 2022 with permission under the guidelines of Royal Society Publishing.