|
SRCD 2019 Biennial Conference
March 21-23, 2019 Baltimore, Maryland Title: Theoretical and Methodological Implications of Associations Between Executive Function and Mathematics in Early Childhood Event: Unpacking Associations Between Children’s Early Executive Function and Mathematics Skills Authors: Tutrang Nguyen, Robert Duncan, Drew Bailey Abstract: Despite consensus on the importance of EF to math achievement in early childhood, its treatment in correlational studies of cognitive development sometimes reflects a lack of agreement about how EF is related to math, either through an underlying construct or through the individual components that comprise the construct. Importantly, hypothesized models make very distinct predictions about the effects of EF interventions on children’s achievement. If factors common to EF components fully account for the relations between children’s EF and math achievement, training the specific components alone is not predicted to influence children’s math achievement. In contrast, if the specific EF components account for the relations between children’s EF and math achievement, the possibility of raising children’s achievement via EF component training is more promising. We investigate whether correlations among EF tasks and math achievement are consistent with the hypothesis that the association between EF and math achievement operates through specific components of EF, through a single reflective latent EF factor, or both. We use data from a nationally representative sample of typically developing children (ages 3-8) across the U.S. and a database of ten peer-reviewed studies on EF and math achievement. First, we present correlations between EF components and math achievement. Second, we conduct factor analyses to estimate the loadings of each EF component onto the latent EF factor. Third, we compare the statistical fit and theoretical implications of a model that allows for an effect of a broad EF factor on math achievement (Model 1), and another that allows for independent effects of EF components on math achievement (Model 2). Finally, in the model that included a broad EF factor on math achievement, we tested whether components of EF had significant residual correlations with math achievement. For the two hypothetical models considered, in general the model with a single EF factor received the majority of support based on statistical fit indices (see Table 1). For the three time points in the ECLS-K dataset, Model 1 was preferred and fit the data well. Additionally, seven of the nine peer-reviewed studies also preferred Model 1 (and one tie). Table 2 shows the path estimates included between each specific component and math. For the ECLS-K dataset, Model 1 showed that latent EF had estimated βs = .85 – .88. Model 2 showed that working memory was the most closely associated component of EF, βs = .44 – .46, although only time point 1 found working memory to have a positive residual correlation beyond the latent EF factor. We find that a single reflective EF factor accounts for most of the EF component-specific associations with math achievement. Thus, our analysis of multiple correlational studies suggest interventions should consider targeting the underlying EF construct rather than component-specific training in order to boost mathematics achievement. Our findings point to the importance of specifying a measurement model for EF in future work. We suggest researchers match the correlational literature to future intervention designs and carefully consider which types of tasks are influencing which component and at which level. Comments are closed.
|
|
