Poster #38 - Verbal labels support rule-based category learning in children

• In Event: 3-206 - Poster Session 13
  In Poster Session: PS 13 Section - Cognitive Processes

Integrative Statement

What are the cognitive consequences of having a simple word for something? Past work suggests that having a label for simple visual features can support category learning in adults (Zettersten & Lupyan, 2018). For example, categories defined by more easily nameable color features (such as “brown” and “red”) are learned more readily than categories composed of color features that are more difficult to name (such as “lavender” and “mustard”, see Fig. 1A). Labels may support category learning by providing compact hypotheses about category-relevant features – it is easier to form a rule-based hypothesis such as “red belongs to category A” than “greenish-yellow-brownish things belong to category A”.

If labels can help adults discover category structure, how might children’s growing lexicon help guide them to learn new categories? To test this question, we investigated whether the nameability of category features would change how easily children (N = 70, mean age = 62 months, range: 48 – 79 months) could form a rule-based category. In one condition (high nameability condition), category exemplars were composed of features children could readily produce a label for (modal naming agreement > 91%), while for the other (low nameability condition), features were more difficult for children to verbalize (modal naming agreement < 47%). The category learning task was framed as a game in which children learned which food (the category stimuli) was preferred by each “alien”. On each trial, children were presented with a color wheel exemplar and asked to “feed” it to the alien that liked it, receiving feedback after their choice (see Figure 1B). Category membership (i.e. which food was preferred by each alien) was perfectly predictable from a single color, e.g. alien 1 always “liked” exemplars containing a brown color-slice. We simultaneously collected data from adults (N = 74) on the same task in order to replicate the previously observed nameability effect (Zettersten & Lupyan, 2018).

Overall, children succeeded at learning to sort exemplars into the two categories (overall mean accuracy = 65.8%, 95% CI = [62.5%,69.2%]), though adults were far more accurate (M = 92.0%, 95% CI = [90.0%, 94.0%]). Accuracy improved across trials for both children (b = .02, z = 3.95, p < .001) and for adults (b = .11, z = 7.55, p < .001). Crucially, children had higher accuracy in the high nameability condition (M = 69.3%) than in the low nameability condition (M = 62.4%, z = 2.02, p < .05), mirroring the accuracy difference between the two conditions observed in adults (high nameability condition: M = 95.7%, low nameability condition: M = 88.3%, z = 3.55, p <.001; see Fig 2).

In sum, children learned rule-based categories better for more nameable color features. These findings could help to explain why vocabulary growth during childhood predicts children’s performance on (putatively non-linguistic) cognitive tasks as well as educational outcomes (e.g., Morgan et al. 2015). Future work will seek to demonstrate causal evidence for label-driven category learning by asking whether learning words for unfamiliar features aids in subsequent category learning.

Authors

• Martin Zettersten, University of Wisconsin-Madison
  Presenting Author

• Catherine Bredemann, University of Wisconsin-Madison
  Non-Presenting Author

• Megan Kaul, University of Wisconsin-Madison
  Non-Presenting Author

• Haley Vlach, University of Wisconsin-Madison
  Non-Presenting Author

• Heather Kirkorian, University of Wisconsin-Madison
  Non-Presenting Author

• Gary Lupyan, University of Wisconsin-Madison
  Non-Presenting Author