Categorical Perception of Topological Relations between Objects

Abstract

In many types of perception, change are easier to spot a when they are categorical - crossing thresholds between blue and green, or between oblique and straight. Here we show that categorical perception applies powerfully to topological relations between two objects. We showed two pairs of simple objects to participants, and asked them to detect potential changes to the distances between the objects across a 1 second delay. Participants (N=15) were far better at detecting changes that crossed categorical thresholds (M=88%), e.g., from touching to intersecting or from intersecting to containing, compared to changes of equal size that did not cross these boundaries (M=73%) (p < .001). With large memory loads, participants appear to rely heavily on abstract and efficient categorical representations of interobject relations. But under conditions of low memory load, this advantage may be hidden to us. When participants were asked to detect changes to only a single pair, performance for categorical changes (M=93%) and non-categorical changes (M=88%) was more similar (though still significantly different, p = .005). A significant interaction between categorical status and memory load (p < .001) confirmed that the effect was far stronger when load was higher. Previous research shows that humans are skilled at encoding and remembering categorical information. Here we show that this broader principle applies to the perception of spatial relations between objects - a key area of visual processing that has received surprisingly little attention. Meeting abstract presented at VSS 2015