Do Multilingual Friendships Predict Reading Comprehension for Elementary Students?

• In Event: Social Network Approaches to Understanding Peer Relationships and Positive Behaviors

Abstract

Introduction: Schools in the United States are becoming increasingly linguistically diverse (August & Shanahan, 2006; Takanishi & Menestrel, 2017). Yet multilingual students continue to face marginalization due to systemic factors and not because multilingualism is itself a risk factor (Takanishi & Menestrel, 2017). Theories of language development that center multilingualism stress how all students bring their full linguistic repertoires to bear in classroom settings (Kramsch, 2002; van Lier, 2002). Research following this framework has identified the importance of multilingual friendships for fostering academic development of multilingual children in after-school settings (Hwang et al., 2022) and in secondary school (Carhill-Poza, 2015). However, research has yet to explore whether multilingual friendships in elementary classrooms might be an important predictor of academic skills.

Hypotheses: We test whether the peer multilingual network size of multilingual students (i.e. number of friendships where two languages are spoken) in Grades 4 and 5 predicts their reading comprehension skills, controlling for English language skills. We hypothesize that multilingual network size will uniquely predict English reading comprehension.

Study Population: Participants (expected N = 700) will be multilingual students in Grades 4-5 whose primary caregivers report speaking at least some of a language other than English at home. The current data (N =23) comes from pilot data of the ongoing study. The sample will be augmented by data to be collected in the fall of 2022.

Methods: Participants completed measures of decoding, vocabulary, morphology, and reading comprehension. Participants also completed a social network survey, where students selected who they “like to hang out with” from a classroom roster. Students then completed a follow-up question asking what language they use when talking to each peer selected, with a five-point response ranging from “All English” to “All another language.” We then calculated the degree centrality for each students’ multilingual friendships by counting the number of friendships where students reported speaking two languages to at least a modest extent. We then normalized multilingual friendship degree centrality by dividing the number of friendships by the possible friendships a student could have in their classroom. To answer the research question, we used a permutation regression approach predicting reading comprehension with number of multilingual friendships as the predictor of interest, while controlling for the other language and literacy measures. The permutation approach is better suited for the network data, which violates the statistical assumption of independence necessary for typical parametric tests. This approach is used in the social development literature (e.g. Kornbluh et al., 2016). All variables were centered to allow for interpretability of coefficient size.

Results: Preliminary results from the regression indicate that multilingual network size does not predict reading comprehension (B = 0.09, p > .05). As mentioned, these preliminary results are based on a limited pilot sample. Results from the preliminary regression analysis are presented in Table 1. Vocabulary was the only significant predictor of reading comprehension (B = 0.59, p < .05). Ongoing analyses will use network analytic methods to account for relational interdependence of students within the same classroom.

Authors

• Andrew Weaver, New York University
  Presenting Author

• Michael J. Kieffer, NYU Steinhardt

• Patrick Proctor

• Elise Capella, NYU