Similarities in Object and Event Segmentation: A Geometric Approach to Event Path Segmentation

Abstract

Events, like objects, can be decomposed into parts. Path, the spatiotemporal trajectory of an object during an event, is the most commonly labeled event feature across the world's languages, provides important social information, and is increasingly central to theories of general event segmentation. However, little is understood about how adults visually segment paths. We apply theories developed for object segmentation to help understand path segmentation. Overall subjects segmented equivalent object shapes and event paths in similar ways following patterns predicted by Singh and Hoffman's (2001) geometric analysis of object parts. There were two notable differences between object and event segmentation: (1) event parsing occurred at points of negative curvature minima and positive curvature maxima as opposed to simply negative curvature minima; and (2) event parsing was more frequent and variable than object parsing. Implications of these results for event perception and categorization are discussed.

Keywords:

• object segmentation
• event segmentation
• event perception

Notes

¹Obviously, the visual system does not treat paths exactly like edges, the direction of motion and changes in speed are both important aspects of paths. We can tell the difference between going into and out of a house. Although direction and speed changes may provide additional segmentation information, the formal congruence between paths and edges allows mathematical analyses from the spatial domain to be applied in the domain of events.

²The rank order analysis with the Monte Carlo simulation was performed instead of a correlation because a correlation requires corresponding values. As subjects were allowed to indicate multiple segmentation locations, establishing correspondence was impossible without running the same subjects on both tasks. We felt that doing that would lead to a bias to make similar judgments in the two tasks; as establishing whether or not the two tasks were related was a central research issue, such a correlation design was problematic.

³For this and subsequent F-tests of differences in variance in individual paths, the estimate for variability with random segmentations was computed by a Monte Carlo simulation of an equal number of segments that could occur with equal probability at any location within each interval.