Executive Functions and Processing Speed: Same or Different?

• In Event: 3-202 - What variation in executive functioning tells us about ADHD: Insights from twin and longitudinal studies

Integrative Statement

Background:
Both executive functions (EF) and processing speed (PS) are constructs that are emerging as important transdiagnostic correlates of diverse developmental outcomes, including learning disabilities, ADHD, and other developmental psychopathologies. However, both constructs suffer from “task impurity” challenges which has led to questions about whether they may be substantially overlapping constructs. Further progress towards understanding the nature of cognitive transdiagnostic risk requires clearly separating these constructs and assessing their shared and unique contributions to various developmental outcomes. In these analyses, we use latent factor models to address the issue of task impurity, and we fit bifactor models consistent with leading theories of both constructs. Using these models, we test the cognitive relationship between EF and PS and the transdiagnostic predictive value of both constructs for dyslexia and ADHD.

Method:
657 children and adolescents ages 8-16 were drawn from the Colorado Learning Disabilities Research Center (CLDRC), which is a population-based sample of twins recruited for reading problems (28% of the sample) and/or ADHD (25%), and typically-developing controls (57%). The CLDRC includes an extensive cognitive, academic, and behavioral battery. Measures of EF included verbal working memory (digit span backwards, counting span, sentence span) and inhibition tasks (stop signal RT, continuous performance test commissions). Measures of PS included simple (simple and choice RT) and complex speed (coding, symbol search, trails, and experimental tasks).

Results:
Bifactor models of both EF and PS were specified (see Figure 1). For EF, a common EF and working memory factor were observed, but inhibition did not retain variance separate from common EF, which is a finding consistent with “unity and diversity” model of EF (Miyake & Friedman, 2012). For PS, a common PS and simple speed factor were observed, but the complex speed factor did not retain variance separate from common PS. For both PS and EF, bifactor models were superior to 1 factor and correlated traits models. To answer the central question of these analyses, we tested the correlation between common PS and common EF, r=.6, p<.001. This correlation was not diminished after controlling for IQ, indicating that the shared variance between common PS and common EF is largely separate from IQ. In models where we tested the competing predictive value of EF and PS for dyslexia and ADHD symptoms, it was the shared variance between common PS and common EF that showed transdiagnostic prediction.

Discussion:
These results show that common PS and common EF are related but distinguishable constructs; more of the variance is unique (56%) compared to shared (44%). Our models as specified do not support the view that PS and EF are essentially overlapping constructs. Nevertheless, there is substantial shared variance between the two constructs and this variance appears to carry the transdiagnostic predictive value of both PS and EF, at least in the case of dyslexia and ADHD. The nature of this shared variance remains unclear, though the current analyses point to speeded tasks requiring EFs as potential markers of transdiagnostic risk.

Authors

• Lauren McGrath, University of Denver
  Presenting Author

• Bruce Pennington, University of Denver
  Non-Presenting Author

• Erik Willcutt, University of Colorado, Boulder
  Non-Presenting Author