Poster #24 - Using the CHILDES database to evaluate the impact of conversational overlap on child language development

• In Event: 3-001 - Poster Session 2 during Breakfast

Abstract

Young children develop language through social interactions with parents and other caregivers. Over conversational turns, parents and children often repeat each other’s words to build on what was just said; such conversational overlap may serve to ratify and acknowledge the partner’s communicative efforts and establish common ground (Clark & Bernicot, 2008). Recent work, utilizing an archival longitudinal corpus from the Child Language Data Exchange System (CHILDES; MacWhinney, 2000), examined the relative impact of maternal vs. child repetition of their partner’s utterances on the child’s subsequent language growth (Che et al., 2018). Using the CHIP command within the CLAN program to compute proportions of utterances with overlapping content (Sokolov & MacWhinney, 1990), the authors examined growth in mean length of utterance (MLU; Brown, 1973), developmental sentence score (DSS; Lee & Canter, 1971), and vocabulary diversity (VOCD; McKee, Malvern, & Richards, 2000) from age 14 months to ages 20 and 32 months (New England Corpus). The main finding was that maternal overlap of the child’s utterances (i.e., repetition of words the child had just said), but not child overlap of the mother’s utterances, predicted language growth over time. The current study aimed to replicate this finding using another, smaller longitudinal corpus from the CHILDES database: the Bates corpus comprising transcripts of children at 20 and 28 months. Due to insufficient numbers of complete sentences, VOCD could not be computed and was replaced with Word Types as a measure of vocabulary diversity. Table 1 presents results of regression models predicting MLU, DSS, Word Types at 28 months using measures taken at 20 months. In line with Che et al. (2018), maternal overlap, but not child overlap, at the earlier time point tended to predict language outcomes at the later time point; however, due to low statistical power, these associations did not always reach conventional levels of statistical significance when control variables (child and maternal MLU at 20 months) were included in the models. The results provided some support for the view that maternal overlap constitutes a form of just-in-time feedback, aligning with other research emphasizing the contribution of maternal responsiveness to communicative bids in promoting language and communicative development (Bornstein et al., 2008; Gros-Louis, West, & King, 2014). Moreover, results indicated that child MLU and DSS at 28 months were predicted by maternal MLU at the earlier time point, but not child MLU, indicating benefits of more complex input for developing sentence structure.

Authors

• Elizabeth S. Che, The Graduate Center and College of Staten Island, CUNY
  Presenting Author

• Maria F. Alarcon, Long Island University

• Patricia J. Brooks, College of Staten Island and The Graduate Center, CUNY