The structure of working memory and how it relates to intelligence in children

Novelty. This study explored the structure of working memory, and its relation with intelligence in 176 typically-developing children in the 4th and 5th grades at school. Different measures of working memory (WM), and intelligence (g) were administered. Confirmatory factor analyses showed that WM involves an attentional control system and storage aspects that rely on domain-specific verbal (STM-V) and visuospatial (STM-VS) resources. The structural equation models showed that WM predicts a large portion (66%) of the variance in g, confirming that the two constructs are separable but closely related in young children. Findings also showed that only WM and STM-VS are significantly related to g, while the contribution of STM-V is moderate. Importance. The structure of WM and its relationship with intelligence has not received enough attention in children. In the present study, we found that our data fitted poorly with a unitary WM model, and with two-factor models that were either modality dependent (distinguishing between visuospatial and verbal components) or modality independent (distinguishing between STM and WM). Our findings indicate that children's WM can be well represented by three components, which distinguishes between a WM component and two storage components relying on domain-specific verbal and visuospatial resources. This result is consistent with previous research on populations of developmental age (e.g., Alloway et al., 2009). In addition, we found that STM-VS (typically involving unfamiliar situations) predicts a unique portion of the variance not explained by active WM whereas the verbal component (i.e., STM-V) (typically involving more familiar material) is less relevant. Our results also confirm that WM predicts a substantial portion of the g variance even when the effect of STM is taken into account. Methods. The method used was particularly appropriate for investigating the structure of WM and the relationship between intelligence and WM. For a start, a large sample of children from 4th- and 5th grades was included. Second, we used a large number of tests of WM tests, which allowed us to test several alternative WM models. Finally, in order to measure general intelligence, we used three tasks (two verbal and one visuospatial task). This allowed us to have a more robust measure of general intelligence which was, for example, not biased toward visuospatial tasks.