2018 Sat Session B 0930

Saturday, November 3, 2018 | Session B, Conference Auditorium | 9:30am

Linking input and vocabulary in infancy to preschool language skills
S. Dailey, E. Bergelson

Differences in early language abilities relate to childhood outcomes (Elliott, et al., 1989; Marchman & Fernald, 2008). Many aspects of children’s language environment are posited contributors to this variability, including input word-type and -token counts (e.g. Hurtado, et al. 2008; Weisleder & Fernald, 2013), reading (Whitehurst, et al., 1988), and words in isolation (Brent & Siskind, 2001). However, the relative effects of such factors have been debated (Swingley & Humphrey, 2017) and may change over the course of development. For example, isolated words may aid infants’ segmentation, while syntactic variability may be critical for later word-learning (Brent & Siskind, 2001; Johnson et al, 2014; Naigles & Hoff-Ginsburg, 1998). Here we examine links between the early language environment and language abilities longitudinally.

We analyzed 44 children’s language input and production in infancy and language skills two years later (M=3;7(14d); followup ongoing, n=18/44). Infant language input was measured through monthly home-recordings from 0;6-1;5; child-directed and child-uttered concrete nouns were annotated for a larger noun-learning project (Bergelson & Aslin, 2017). For present analyses, we averaged the proportion of reading and short-phrases (<3-word utterances, a proxy for words in isolation and at edges) across all months for each subject. Infant vocabulary (age 1;5) was measured using the MCDI (Fenson, et al., 1994) and child word-types from home- recordings. Preschool language abilities were measured with QUILS (Quick Interactive Language Screener; Golinkoff, et al., 2017), which combines processing, syntax, and vocabulary.

We first examined infant vocabulary and preschool language abilities. QUILS scores at 3;7 positively correlate with both infant vocabulary measures at 1;5 (MCDI: Pearson’s R=0.63, p=0.007; child noun-types: Pearson’s R=0.71, p=.001), providing the first longitudinal validation of this new preschool language measure.

We then predicted QUILS scores from the proportions of reading and short-phrases in infants’ input. We find that short-phrases are negatively correlated with QUILS scores (Pearson’s R=-0.55, p=0.02), while reading utterances are positively correlated with them (Pearson’s R=0.5, p=0.03). When compared to a baseline model predicting QUILS using child word-types at 1;5, adding proportion of short-phrases in the input significantly improves model fit (F(1,14)=11.71, p=.001), while adding proportion of reading utterances does not (F(1,14)=0.09, p=.31). Thus, short-phrases in children’s early environments predict language scores at 3;7 above and beyond productive vocabulary at 1;5.

Summarily, we find evidence that both vocabulary and language input in infancy predict overall language abilities two years later. While the role of reading in language development is well-established (Debaryshe, 1993), the negative relationship between short-phrases and language abilities is a new result. Intriguingly, while isolated words are linked to word learning at the word level (Swingley & Humphreys, 2017), having larger proportions of such input at the child level may have deleterious effects on later language. This underscores the role of syntactic diversity in language learning (Naigles & Hoff-Ginsberg, 1998). These results point to complex input-output functions in early language, highlighting both the conservation of early skills at later time-points and the role that language environment in infancy plays in language abilities years later.