In search of better practice in executive functions assessment: methodological issues and potential solutions

The multi-component nature of executive functions (EF) has long been recognized, pushing for a better understanding of both the commonalities and the diversity between EF components. Despite the advances made, the operationalization of performance in EF tasks remains rather heterogeneous, and the structure of EF as modelled by confirmatory factor analyses (CFA) is still a topic of debate (Karr et al., 2018). The present work demonstrates these two issues are related, showing how different operationalizations in task-based performance indicators impact the resulting models of EF structure with CFA. Using bootstrapped data from 182 children (8-12 years old) and nine EF tasks (tapping inhibition, working memory and cognitive flexibility), we first show improved model convergence and acceptance when operationalizing EF through single tasks’ scores (e.g., incongruent trials, Flanker task) relative to difference scores (e.g., incongruent minus congruent trials, Flanker task). Furthermore, we show that reaction times exhibit poor model convergence and acceptance compared not only to accuracy, but also drift rate. The latter, a well-known indicator in drift-diffusion models, is found to present the best psychometric properties to model EF with CFA. Finally, we examine how various operationalizations of performance in EF tasks impact CFA model comparison in the assessment of EF structure and discuss the theoretical foundations for these results.