Poster #66 - Language and shape co-occur together: Linguistic and extralinguistic labels in relation to children's shape bias

• In Event: Poster Session 17

Abstract

Introduction. The tendency for individuals to generalize a novel label to a novel object based on shape is known as the shape bias (Baldwin, 1992). Emerging around 2-3 years of age, the shape bias is a highly replicable phenomena, yet there is disagreement in the literature as to what drives this bias. The Attentional Learning Account (ALA) posits that through the experience of learning new words, children become attuned to the perceptual features of objects that allow them to more easily categorize these objects into discrete categories (Samuelson & Smith, 2006). On the other hand, the Shape-as-Cue Account (SCA) argues that the shape bias is not the result of associative learning between shape and words, rather that shape is a reliable way to categorize all objects (Bloom & Gelman, 2000). In this project, we examined whether children generalize based on shape when presented with various types of labels for objects, such as words, animal sounds, and machine sounds. In line with ALA, we predicted that children would show a stronger shape bias with words than with animal and machine sounds. That is, children’s generalization of shape would be in line with their learning history of associations between words, sounds, and objects.

Methods. In Experiment 1, 2–4-year-old children (N=53) were presented with a classic shape bias task (Figure 1) and assigned to one of two conditions; a linguistic condition, where a novel object was paired with a novel linguistic label (“wug”), and an extralinguistic condition, where a novel object was paired with a novel, machine-like noise. In Experiment 2, we replicated these two conditions and added a third label condition: an animal sounds condition. This third condition was added to rule out the possibility that role communicative intent drove the results observed in Experiment 1.

Results. We first determined whether children had a shape bias. According to a binomial probability calculator, 9 or more shape choices (out of 15) was significantly more than would be predicted by chance, and thus we used this criterion for determining a shape bias. To examine whether children showed a stronger shape bias when presented with novel words (63% of sample) vs. novel machine-like sounds (42% of sample), we conducted an independent sample t-test on the number of children with a shape bias. The results revealed that more children in the novel linguistic condition had a shape bias, p = .046. Experiment 2 data collection is ongoing (N = 35, planned N = 75) and will be completed by the conference.

Conclusions. The results of Experiment 1 support the ALA over the SCA: Children had a stronger shape bias when presented with words, suggesting that language history contributes to the degree to which children use shape to categorize objects. Indeed, this work implies that children’s generalization of object features is dynamic depending on contextual factors, such as the linguistic and auditory context of in-the-moment of learning.

Authors

• Kaitlynn Ellis, University of Wisconsin - Madison
  Presenting Author

• Benjamin Barmore, University of Wisconsin - Madison

• Nathan J Simon, University of Wisconsin - Madison

• Melina Lauryn Knabe, University of Wisconsin - Madison

• Haley A Vlach, University of Wisconsin - Madison