Date of Award

Summer 2023

Degree Type

Open Access Dissertation

Degree Name

Psychology, PhD

Program

School of Social Science, Politics, and Evaluation

Advisor/Supervisor/Committee Chair

Andrew A.R.A Conway

Dissertation or Thesis Committee Member

Kathy Pezdek

Dissertation or Thesis Committee Member

Gabriel I. Cook

Dissertation or Thesis Committee Member

Kristof Kovacs

Terms of Use & License Information

Terms of Use for work posted in Scholarship@Claremont.

Rights Information

© 2023 Kevin P Rosales

Keywords

attention, executive functions, latent variable modeling, network modeling, psychometrics, Working memory

Subject Categories

Cognitive Psychology | Psychology

Abstract

Beginning in the 1970s, a great deal of research in cognitive psychology, developmental psychology, psychometrics, and cognitive neuroscience has investigated the structure and function of working memory (WM), defined as the ability to actively maintain and manipulate information in the service of complex cognition (Baddeley & Hitch, 1974). It is well established that WM is a limited capacity system and individual differences in WM capacity are strongly associated with important cognitive abilities and outcomes, such as general intelligence (Engle et al., 1999) and academic achievement (Swanson & Berninger, 1996; Ramirez et al., 2013). For this reason, WM is a central component in most general theories and models of cognition. However, over the years, different researchers have proposed different definitions of WM. This is problematic because researchers who adopt different definitions of WM also tend to administer different kinds of tasks to measure WM capacity, which has produced a pattern of inconsistent results reported throughout the literature. This inconsistency has led to a lack of a consensus in the field regarding how to measure WM capacity and how to determine the “best” psychometric model of the structure of WM capacity. If we look to the most prominent contemporary theories of WM, both cognitive and psychometric, we can identify several different components of WM function that are thought to contribute to individual differences in WM capacity. These include attention control (Engle, 2002), verbal and spatial temporary memory storage (Kane et al., 2004), and episodic memory retrieval (Unsworth et al., 2014; Oberauer, 2009). Though these components have been shown to contribute to variation in WM capacity, there currently is not a comprehensive psychometric model of WM that includes all of these components. Moreover, most of the research on individual differences in WM capacity has been conducted using traditional latent variable modeling approaches (factor models), which are based on problematic assumptions (Borsboom et al., 2003). More recently, network analysis has emerged as an attractive alternative psychometric modeling approach to study individual differences in cognitive abilities (Kan et al., 2019). Network analysis does not rely on the same problematic assumptions required by latent variable models, and it is more compatible with recent theories of intelligence, namely, Process Overlap theory (POT) (Kovacs & Conway, 2016). POT proposes that broad cognitive abilities reflect multiple cognitive processes that are sampled in an overlapping manner across a range of cognitive tasks. This theoretical framework aligns well with network modeling where abilities are represented as an inter-connected network of multiple cognitive processes.Across two studies in this dissertation, we (1) compared network models to traditional latent variable models of WM capacity, with both types of models designed to include multiple components, namely, attention, verbal storage, spatial storage, and episodic memory retrieval and (2) tested the predictive validity of the models by estimating the correlation between WM capacity and fluid intelligence. The results show that a network model of WM fits the data just as well as a latent variable model, as predicted. However, we did not support the hypothesis that the network model of WM predicts fluid intelligence equally well as the latent variable model. Taken together, the current studies provide new insights into the psychometric structure of WM using the novel technique of network modeling. It is shown that a four-component network model of WM capacity is an accurate and comprehensive depiction of WM. These results help to integrate cognitive models and psychometric models of WM, which is an important contribution to the field and has implications for research and practice in clinical and educational settings where measuring WM capacity effectively, and interpreting WM test scores properly, is of the utmost importance.

ISBN

9798380478601

Download

Share

COinS