Search

Background: Decades of research demonstrating the critical role of executive function (EF) and other regulation-related skills on children's short- and long-term outcomes has generated interest in measuring and cultivating these critical skills in early childhood and elementary school children (e.g., Bull et al., 2008; Moffitt et al., 2011). Most prior work on EF operationalizes, measures, and analyzes outcomes at the child level, meaning the analysis typically focuses on individuals' skill development without accounting for the dynamic, social, interactive world in which skills develop (Bailey & Jones, 2019). For example, we still don't have much basic information on variation in EF between individual children in classrooms, variation between classrooms, or on compositional features that influence EF skills over time. Designing effective interventions and aligned, relevant measurement tools demands more knowledge about the factors at multiple levels of children's ecologies (i.e., both individual characteristics and classroom phenomena) that are linked to variation in children's EF skills over time. Research Questions: In this study we address the following research questions: 1. How much variation in EF scores exists (1) between schools, (2) within schools (i.e., between classrooms), and (3) within classrooms (i.e., between children) at the beginning and end of one school year (fall and spring)? What classroom and individual factors predict variation in children's EF in the spring? 2. What predicts children's spring EF scores? Do aggregate (i.e., classroom-level) EF skills predict children's outcomes over and above individual EF scores? Setting and Participants: Data for this study come from a randomized controlled trial of the Social, Emotional and Cognitive Understanding and Regulation in education (SECURe) program, an intervention focused on improving the SEL skills of elementary-aged children. Six schools in Phoenix, AZ, were randomly assigned to the SECURe intervention or to a control condition (three schools to each). All six schools were designated as schoolwide Title I schools. Most children were Hispanic or Latino/a (see Table 1). Direct assessments were conducted with all children in three randomly selected classrooms per grade, resulting in a sample of 518 direct assessments from 37 kindergarten and grade 1 classrooms and 572 direct assessments from 30 grade 23 classrooms. Children were assessed using two direct assessments of EF skills from the Woodcock-Johnson III Tests of Achievement (attention control and working memory) and an assessor report of children's attention and impulsivity during the assessment period using the Preschool Self-Regulation Assessment Assessor Report (PSRA-AR) in the fall and spring of that year. Data Collection and Analysis: To explore the first research question about the amount of variation within and between classrooms, we examined intraclass correlations (ICCs) for each fall and spring outcome in multilevel regression models, first with random intercepts for each school and then with random intercepts for each classroom. To examine predictors of variation in levels of EF, we constructed a taxonomy of seven models for each of our primary outcomes. The first is a null model with only the primary outcome and classroom random intercept; the second model adds child gender. The next three models add classroom-level variables (described below). The final two models add school-level variables--a dummy variable for the SECURe intervention in model 6, and school fixed effects in model 7. Classroom-level EF was operationalized in three ways for each measure. First, we created a leave-out classroom mean, calculated by averaging baseline scores for children in each classroom without the child's own score (i.e., leaving out that child's score). To identify children at the upper and lower quartiles of the distribution, we calculated a variable capturing the number of children in each classroom scoring equal to or below the grade-level 25th percentile on each measure, or equal to or above the grade-level 75th percentile on each measure in each grade, meaning that one cut score existed for each grade level, across all schools. All analyses were done in grade bands (kindergarten and grade 1; grades 2 and 3). To address the second research question, we present a set of multilevel regression models following the same taxonomy as above, predicting spring outcomes with (1) individual level predictors--child baseline scores and gender, (2) classroom level predictors, and (3) school-level variables--the SECURe intervention and school fixed effects. Results: Our findings indicate greater variance in children's EF within schools than between, with very little or no variance arising from differences between schools. Though we observed greater variance within classrooms (i.e., between children and therefore assumed to derive from individual characteristics and experiences) than between them, a substantial amount of variation in EF scores appears in our data to arise from differences between classrooms (see Table 2). In several cases, the leave-out classroom mean accounted for a similar or greater amount of variance than did children's individual baseline scores. As expected, children's individual fall scores predicted spring outcomes across all models. A key finding from this study is that classroom-level variables, particularly the leave-out classroom mean, were also significant predictors of children's outcomes, and in many cases were stronger predictors than were children's own baseline scores. The relatively consistent pattern of results for the leave-out classroom mean and the magnitude of the coefficient compared to the individual scores suggests that the skills of children's peers constitute, in some cases, a stronger predictor of their spring EF skills than do their own individual skills at the beginning of the school year. Overall, school-level factors were not significant predictors of children's spring outcomes. However, random assignment to the SECURe intervention was associated with increases in attention control for second- and third-graders. See Tables 3-5. Conclusions: Findings from this study suggest that an ecological approach to intervention design, measurement, and analysis would provide richer perspectives on the role of settings in student's development than the largely dominant individualized approaches. The presentation will discuss implications for intervention and assessment in school and classroom contexts.