Using Behavioral and Brain Measures of Word Learning to Understand the SES Vocabulary gap in Grade School

• In Event: 1-208 - The Neural Basis of Word Learning

Integrative Statement

Background. Two decades of research has uncovered complex environmental factors that put young children from low socioeconomic status homes (LSES) at risk for slow vocabulary growth compared with their higher SES (HSES) peers (eg., Hoff, 2013). Why the vocabulary gap increases over the school years is less well understood, however previous studies indicate that SES differences in grade-schoolers’ ability to learn new words from written context may play a role (Maguire, et al., 2018). Here we use EEG, standardized assessments, and an experimental word learning from context task to investigate if and why LSES children have more difficulty using known language to infer the meaning of novel words. We focused on reading and vocabulary, skills that are often difficult for LSES children, and two ERP components. The P2 indexes early reading processes including phonological encoding and lexical selection (Coch & Gabriela, 2015; Indefrey, 2011) and the N400 indexes semantic retrieval (Kutas & Hillyard, 1980) and word learning (Mestres-Misse, et al., 2010).

Methods. Twenty-five LSES 12-15-year-olds (based on free or reduced lunch qualifications) and 25 age and sex-matched peers (See Table 1) completed vocabulary and reading assessments. During the word learning task they wore a 64-electrode EEG cap while silently reading 100 sentence triplets with a target novel word in the sentence-final position. After each triplet, participants reported the novel word’s meaning, if possible.

EEG Analysis. We studied ERP responses at two points: 1. all familiar words across the course of the sentence and 2. each presentation of an unknown word as it is learned. Only trials in which children identified the target word were included in the EEG analysis. EEG data cleaning and epoching followed standards in the field (EEGLab Tutorial). We used EEGLAB’s Monte Carlo permutation analysis, which accounts for multiple comparisons, to statistically compare ERP responses between LSES and HSES children.

Results.
Behavioral. Mediation analysis revealed that family income predicted word learning through the mediating factor of vocabulary, even when covarying for reading comprehension, decoding, and age, b=.009, SE=.006 CI=[.001 .025].

ERPs. At each familiar word leading up to the target, we observed significant SES differences in the P2 amplitude, with LSES children exhibiting smaller amplitudes than MSES children indicating greater difficulty with reading and lexical selection. No SES differences in the N400 were found for known words. For the target word, similar to adults (Mestres-Misse et al., 2010), HSES children exhibited a significant attenuation of the N400 across trials, p’s<0.05. LSES children did not. See Figure 1.

Conclusions
These findings support the claim that differences in the ability to learn new words from context is a contributing factor to ongoing vocabulary differences between LSES and HSES children during the school years and elucidate how differences in reading ability, undetectable by behavioral measures, may contribute to differences in learning outcomes. The ERPs highlight how difficulties in early components of reading and lexical selection (P2) may snowball over the course of the sentence and ultimately affect word learning (N400). Implications for education and interventions will be discussed.

Authors

• Mandy J Maguire, University of Texas at Dallas
  Presenting Author

• Yvonne K Ralph, University of Texas at Dallas
  Non-Presenting Author

• Sonali Poudel, The University of Texas at Dallas
  Non-Presenting Author