Phases of systematic brain processing differentially relate to cognitive constructs of attention and executive function in typically-developing children: a latent variable analysis
Date
2017
Authors
Taylor, Brittany Kristine, author
Gavin, William, advisor
Davies, Patricia, committee member
Seger, Carol, committee member
Shomaker, Lauren, committee member
Journal Title
Journal ISSN
Volume Title
Abstract
The series of studies presented in this dissertation examines the complex interrelationships between brain measures, cognitive abilities, and simple behaviors in typically-developing children. Much recent research has been dedicated to understanding the interaction between neural processing and behaviors across development. However, the field continues to rely on simplistic statistical approaches (e.g., correlations, t tests, ANOVAs), which 1) are unable to simultaneously examine multiple interrelationships among variables of interest, and 2) are easily confounded by sources of measurement error. The result is weak relationships between brain and behavioral measures. In this series of studies, we progressively demonstrate how more sophisticated statistical approaches, namely structural equation modeling (SEM) techniques, can be utilized in order to improve researchers' ability to detect brain-behavior relationships in children. All three of the present studies utilize event-related potential (ERP) and behavioral data collected from a sample of typically-developing children ages of 7- to 13-years-old during two separate sessions. In Study 1, we explore the interrelationships between the E-wave component of an ERP, two trait behavioral measures of attentional processing, and simple reaction time (RT) measures during the ERP task. Whereas simple bivariate correlations indicated that the E-wave and RT only shared 7.9 – 9.6% of their variance, a latent variable approach using E-wave and trait attention measures successfully predicted 47.7% of the variance in RT. However, the predictive coefficient from brain-to-behavior was still weak (β = .23), suggesting that there may be neural influences in addition to the E-wave that contribute to the variance in RT. Thus, in Study 2 we elaborated on this model and explored whether the full time-course of an averaged ERP could be conceptualized as a sequence of phases that represents stimulus-to-response decision-making processes. Specifically, we tested a latent variable path model in which one ERP component predicted the next in chronological order, with the full stream of neural processing ultimately predicting RT during the task (N1 → P2 → N2 → P3 → E-wave → RT). Age served as a control variable on each phase of processing and on RT. Results indicated strong predictive relationships from one component to the next (β's = .59 - .86), with the full stream of processing significantly predicting RT (β = .45). The model was fully-mediated, underscoring the importance of the full time-course of the ERP for understanding behaviors during the task. In addition, there were significant age effects on the N2, P3, and RT latent variables (β =.28, -.48, & -.42 respectively). Given the nature of path analyses, the findings suggested that "age" was likely a multifaceted construct representing maturation within multiple domains of cognitive or motor functioning. Study 3 explored the differential relationships between two developmentally-sensitive cognitive constructs and each of the phases of neural processing, effectively replacing "age" with more substantive definitions of maturational effects in the model. The two cognitive constructs captured aspects of attention and executive function processing. Indeed, the findings indicated that each phase of neural processing was differentially influenced by each of the two cognitive constructs. The data suggested that children with better, more matured abilities within a specific cognitive domain tended to have smaller amplitude ERP components from the N1 through the P3, and larger amplitude E-wave components. Conceptually, children with more matured cognitive abilities were able to process the ERP task more efficiently (or with less effort), and engaged in greater anticipatory processing leading to the task behavior when compared to children with less matured cognitive abilities. Of note, the full model did still significantly predict RT during the task, and to a much greater extent than was found in Study 2 (β = .92). The series of investigations in this dissertation demonstrate the utility of SEM approaches for understanding brain-behavior relationships in typically-developing children. Namely, the studies showed that 1) latent variable approaches are helpful in reducing measurement error in ERP and behavioral data, which may impede the detection of brain-behavior relationships when using more simplistic statistical approaches; 2) conceptualizing the full time-course of an ERP preceding a task behavior is not only helpful, but necessary to successfully predict behaviors; and 3) we can further elucidate unique influences of maturation on neural processing within multiple cognitive domains when we embrace advanced statistical approaches like SEM. Implications of the findings and import to the field are discussed in the final chapter.