Poster #40 - Adults’ naïve theories of word learning affect their age of acquisition estimates for words.

• In Event: Poster Session 05

Abstract

Introduction: A key question in the study of children’s language learning is when children learn various words. Researchers have developed several ways to measure a word’s age of acquisition (AoA). The two most common methods involve asking adults: using checklists to ask parents which words their child currently says and using retrospective judgments wherein adults indicate when they think they learned various words. How do these methods compare to ages of acquisition obtained by directly observing what words children can say? Do different measures reveal different biases?

Methods: We compiled several previously published AoA measures: “objective” AoA derived from children’s picture naming (Morrison et al., 1997), parental-report AoA estimates from WordBank (Frank et al., 2017), and retrospective AoA from adult raters (Kuperman et al., 2012). As an additional baseline, we collected “naïve” adult judgments of word knowledge of hypothetical 18-month- and 24-month-olds (N = 87). We included several AoA predictors: word concreteness (Brysbaert et al., 2013) and frequency in child-directed speech (CHILDES; MacWhinney, 2000), which are known to predict AoA (Braginsky et al., 2019). We also included several novel predictors: adult judgments (N = 127) of how strongly each word is associated with preschoolers (“preschoolness”; similar to the “babiness” measure in Perry et al., 2015). We also collected ratings of how helpful adults thought words would be for a preschooler to know (“helpfulness”; N = 106).

Results: Analyses were conducted with 115 words for which all age of acquisition measures existed. These were limited to largely concrete nouns because “objective” AoAs were taken from a picture-naming task. Objective AoA was most strongly correlated with parent-report AoA (Figure 1). AoA measures based on adult ratings were more strongly correlated with each other than with objective AoA. Next, we examined whether different AoA measures were differentially related to the four lexical predictors: frequency, concreteness, preschoolness, and helpfulness. Concreteness and frequency were consistently predictive: more concrete and more frequent words were judged to be learned earlier. Preschoolness and helpfulness were not consistently predictive: higher preschoolness and helpfulness ratings were associated with later AoAs as assessed through children’s language production. In contrast, adults judged words they associated with preschoolers (e.g., slide, crayon) and those judged to be most helpful (e.g., mitten, spoon) as having somewhat earlier AoAs (Figure 2).

Conclusion: Whether adults report on their own child, a hypothetical child, or themselves as a child, they appear to share intuitions about what words are learned early vs. later. Some of these intuitions, such as sensitivity to concreteness and frequency, are correct in that they predict children’s measured language production. Other intuitions appear incorrect, such that children learn words associated with preschoolers earlier than more “adult” words, and that more “helpful” words are learned earlier. Although adult-report AoAs address the limitations of measuring children’s language production, researchers should exercise caution: adult-report measures can broadly capture which words are learned early but are susceptible to adults’ naïve theories about children’s word learning. These intuitions might bias AoA norms, yet a better alternative is not currently available.

Authors

• Melina Lauryn Knabe, University of Wisconsin - Madison
  Presenting Author

• Christina Schonberg, University of Wisconsin - Madison

• Haley A Vlach, University of Wisconsin - Madison

• Gary Lupyan, University of Wisconsin-Madison