Over the Reading Rainbow: Frequency of Color and Spatial Words in Storybooks and Child-Directed Speech

• In Event: Caregiver-child interaction

Abstract

Young children are exposed to many sources of language input, including spoken and written language during shared book reading. The language in children’s books is higher in both lexical diversity and grammatical complexity relative to child-directed speech (CDS; Montag et al., 2015; Montag, 2019). However, it remains unknown whether the specific types of vocabulary children hear vary across these language sources. Do children learn about some words primarily from storybooks? As a first step towards answering this question, we investigate two categories of words that are known to have protracted developmental trajectories – color and spatial words. Specifically, previous research has shown that young children first learn that color words (e.g., blue/red) belong to a class but refine their mappings to become more precise over time (Sandhofer & Smith, 1999). Similarly, children first appear to comprehend and produce spatial words (e.g., by/high) in highly constrained contexts before using them more flexibly (e.g., Hund & Plumert, 2007). Here, we compare the frequency of color and spatial words between text from storybooks and CDS.
We calculated the frequency of the first-learned color and spatial words in corpora of children’s storybooks and CDS. These words were 8 color words and 25 spatial words listed on the MacArthur-Bates Communicative Development Inventory, a normed list of words produced by most children before 2.5-years-of-age (MBCDI; Fenson et al., 1994). Storybook word frequency came from a corpus of 100 picture books that caregivers may read to young children (generated by Montag et al., 2015). CDS word frequency came from CHILDES (MacWhinney, 2000; Baath, 2010). Because the corpora varied greatly in the total number of word tokens they included, we divided each word’s reported raw frequency by the sum of all tokens in each corpus (adding up the raw frequency of all words), controlling for corpora size.
We then used mixed-effects regression to predict the frequency rate of each word from its type (color or spatial), corpus (storybooks or CHILDES), and the interaction between type, corpus, and age of acquisition (AOA, based on MBCDI norms). The model also included a random item intercept to account for variance associated with individual words. There was a significant association between AOA and frequency, X2(1)=8.17, p=.004, such that earlier learned words were more frequent than later learned words. There was an interaction between corpus and type, X2(1)=4.21, p=.040. Follow-up analyses revealed that both color words, X2(1)=23.47, p<.00001, and spatial words, X2(1)=16.20, p<.00001, were significantly more frequent in storybooks than in CDS.
Taken together, the results extend prior work showing that storybooks contain more diverse vocabulary than CDS, by revealing that specific classes of vocabulary, namely color and spatial words, also appear more frequently in storybooks than in CDS. Notably, the average frequency rates for words examined here were low (range: .000001-.01), suggesting that additional factors like gestures and visual regularities may be needed to support learning. Future work should explore what multimodal cues caregivers use to supplement color and spatial word learning during shared book reading and in other contexts.

Authors

• Megan G Lorenz, Augustana College
  Presenting Author

• Lynn K Perry, University of Miami