Variation in Home Language Input is Linked to Predictive Language Processing

• In Event: 1-127 - Resources Talk: A Closer Look at SES, the Home Environment, and Language-Related Outcomes

Integrative Statement

Individual differences in children’s language processing and vocabulary size correlate with language input and socioeconomic status (SES), suggesting that environmental factors play an important role in language development (Hoff, 2003; Fernald, Marchman & Weisleder, 2013). Specifically, language processing efficiency mediates the link between language input and vocabulary size (Weisleder & Fernald, 2013). Thus, disparities in children’s input and in their language processing abilities may interact to drive divergent learning trajectories.

We hypothesized that disparities in language input could also influence prediction (i.e., the ability to anticipate upcoming information during language processing). Prediction is a proposed language learning mechanism (Elman, 1990; Dell & Chang, 2014) and children vary in the extent to which they predict (Mani & Huettig, 2012). Thus, given findings linking language input and language processing (e.g., Weisleder & Fernald, 2013), we hypothesized that language input also supports children’s abilities to predict.

We recruited 28- to 32-month-old children (N=34) from families with varying SES. Language input, which was significantly correlated with SES (r(32)=0.43, p=0.01), was used to create a High-Input group and a Low-Input group, based on a median split (Table 1). We used two eye-tracking tasks to evaluate prediction. In the first task, children viewed images (e.g., a cookie, a book) and heard Predictable sentences with informative verbs (e.g., Eat the cookie), and Neutral sentences (e.g., Look at the cookie). In the second task, toddlers viewed images (e.g., one cookie, two apples) and heard Predictable sentences with informative number marking (e.g., There is the nice cookie), and Neutral sentences (e.g., Look at the nice cookie).

We analyzed children’s looks to the target image (e.g., cookie) during a time window from 300 ms after the onset of the informative verb or number marking until 1000 ms after the target noun onset. If children predict the noun, then they should have greater target looks for Predictable sentences than for Neutral sentences. To evaluate the timing of effects, we analyzed looks within 100-ms time-bins using paired sample t-tests, and used FDR to correct all p-values for multiple comparisons (Benjamini & Hochberg, 1995).

In the first eye-tracking task, we found a significant effect of condition from -100 to 400 ms from noun onset for High-Input children (ps < 0.05), but no significant effects for Low-Input children. In the second task, we found a significant effect of condition from -200 to 800 ms from noun onset for High-Input children (ps < 0.05), but again found no significant effects for Low-Input children. Thus, results revealed that High-Input children predicted the upcoming noun – rapidly looking to the target image before it was named – but Low-Input children did not do so reliably (Figure 1).

Together, these findings suggest that children who receive more language input are more likely to predict during language processing. If prediction is a language learning mechanism, then early differences in prediction, in combination with other factors, may play a role in children’s divergent learning trajectories. Discussion will focus on the range of possible sources of disparities in children’s ability to predict.

Authors

• Tracy Reuter, Princeton University
  Presenting Author

• Casey Lew-Williams, Princeton University
  Non-Presenting Author