Poster #39 - Parents' Descriptive Input for Different Domains in a Semi-naturalistic Setting

• In Event: Poster Session 19

Abstract

A common approach to studying word learning assumes that children learn all nouns in much the same way -- that children learn the word “strawberry” in the same way that they learn the word “frog” or “farmer.” However, children show very different developmental trends and very different knowledge of different domains of words (Gelman, 1998; Djalal et al., 2017). In this study, we examine how the language input is structured in three domains of words (i.e., animals, foods, people) that may lead to differences in acquisition and knowledge.

Participants were 32 children (n = 32; 55.09 months; 16 females) and their mothers. We selected the Individual Differences corpus (Gelman et al., 2014) in the CHILDES database (CHILDES; MacWhinney, 2000). This corpus was selected because parents and children viewed three different categorical domains: animals (e.g., frog), foods (e.g., strawberry), and people (e.g., farmer). Data were analyzed using t-tests corrected for multiple comparisons.

Parents’ utterances about pictures depicting people were highly distinguishable from their utterances about animals and foods. Parents used a lower proportion of statements when talking about people, specifically descriptive statements, compared with animals (t(21)=3.442, p=0.002) and foods (t(21)=2.997, p=0.007). The types of questions parents used also varied. Parents used a higher proportion of non-descriptive (fill-in-blank) questions than descriptive (yes/no) questions for people (t(21)=3.999, p<0.001), but not for the other two domains. Parents used a lower proportion of labels for people compared with animals (t(21)=6.626, p<0.001) and foods (t(21)=7.236, p<0.001), but no significant difference between animals and foods.

The features that parents talked about also differed by domain, with a higher proportion of perceptual features than conceptual features discussed for pictures depicting people (t(14)=7.785, p<0.001) and a lower proportion of conceptual and taxonomic features for people, compared with animals (C: t(21)=4.128, p<0.001; T: t(21)=2.498, p=0.021) and foods (C: t(21)=2.841, p=0.010; T: t(21)=3.903, p<0.001).

One possibility is that these differences were due to differences in familiarity – that children were more familiar with the food and animal words than the people words. To examine this, we compared the average age of acquisition for each word in the three domains. We found no differences between the age of acquisition between the animal, food, and people words (F(2, 58)=2.490, p=0.093). We next divided the pictures within each domain into earlier vs. later acquired words. There was an interaction between children’s age of acquisition of referents’ labels and domains on parents’ use of labels (F(2,42)=5.708, p=0.011), and perceptual features (F(2,42)=3.614, p=0.046). Parents talked about a higher proportion of labels (t(21)=2.250, p=0.035) and perceptual features (t(21)=2.774, p=0.011) for later-acquired than for early-acquired food words. Another possibility, and subject of the discussion, is that people referents have a more malleable category structure and are more similar to each other than the referents within the other two domains.

Altogether, the results indicate that parents talk about, and thus structure, the learning environment in different ways for different domains of words. These differences in language input, as well as category structures, affect lexical pathways supporting language acquisition.

Authors

• Erjing Zhang, University of California - Los Angeles
  Presenting Author

• Catherine Sandhofer, University of California - Los Angeles