Competition between topology and shape in children’s representations of object kinds

• In Event: Poster Session 10
  In Poster Session: PS 10 Topic - Cognitive Processes

Abstract

Children as young as 2 and adults extend the labels of novel objects to other objects that share the same shape, but not to objects that share other surface features like color and texture, suggesting that shape holds a privileged role in representations of object kinds – a phenomenon known as “shape bias” (Landau et al., 1988; Diesendruck & Bloom, 2003). Topological properties of objects (i.e. whether an object has a hole) are fundamental to object recognition; infants and adults are faster to recognize changes in objects’ topological properties compared with shape (Chen, 1982; Chien et al., 2012), and the ability to track multiple objects is disrupted when objects change topology but not shape (Kibbe & Leslie, 2016; Zhou et al., 2010), suggesting that topological class plays a structural role in representations of objects. We hypothesized that topological class also may play a fundamental role in the formation of representations of object kinds.
Sixty-six 2-7-year-olds (M=5.0) participated in Experiment 1. On each of the four trials, the experimenter showed children a novel object (the standard) and gave it a novel label (e.g. “this is a toma!”). The experimenter then showed children three other objects: one with the same shape and different topology as the standard, one with the same topology and different shape, and a distractor object (see Figure 1a). The experimenter asked children to choose which of the three objects shares a label with the standard (e.g. “Which of these is also a toma?”). In control Experiment 2, we tested 21 2-7-year-olds (M=5.01 years) using the same procedure as Experiment 1, except that holes were replaced by gray patches, equating the topology across all objects but changing the surface structure (figure/ground relation) of the objects (Figure 1b).
For each child, we computed a Bias Score (proportion trials shape-match chosen – proportion trials topology-match chosen) (children selected the distractor object on <13% of trials). Bias scores closer to 1 indicate shape bias, closer to -1 indicate topology bias, and closer to 0 suggest no bias. In Experiment 1, children’s Bias Scores were not significantly different from 0 (M=.09; t(65)=1.20, p=.235, Bayes factor=5.13 in favor of null). By contrast, in Experiment 2, Bias Scores were significantly higher than 0 (M=.69; t(20)=6.34, p<.001, Bayes factor=6493.5 against the null), replicating the classic shape bias effect. Bias Scores were significantly higher in Experiment 2 compared with Experiment 1 (t(85)=4.05, p<.001, Bayes factor=119.1 against the null; see Figure 2). Children’s age in days was not correlated with Bias Scores (Experiment 1: r(65)=.17, p=.180; Experiment 2: r(20)=.09, p=.699).
These results suggest that topology may compete with shape in children’s extension of novel nouns to like kinds. We suggest that the topological properties that are fundamental to object recognition and tracking also play a fundamental role in representations of object kinds across development.

Authors

• Praveen Kenderla, Boston University
  Presenting Author

• Sung-Ho Kim, Ewha Womans University
  Non-Presenting Author

• Melissa M. Kibbe, Boston University
  Non-Presenting Author