Relationships between common factors of fluid cognition

Understanding the source of the positive correlation between working memory capacity (WMC) and fluid intelligence (Gf) could benefit those who struggle in daily life. The binding hypothesis posits that WMC and Gf are jointly limited by the number of bindings in working memory (WM), but the mechanisms of binding are unclear. The processing speed theory proposes that fluid cognition declines in older age due to a processing speed bottleneck. However, components of processing speed may differentially relate to fluid cognition across adulthood. This thesis advances understanding of the WM-Gf overlap using methods to partition variance of underlying mechanisms.
Chapter 2 investigated the Three-Embedded Components (3Cs) model of WM using bi-factor analysis and psychometric network analysis to investigate the partial correlations between indicators of the 3Cs, WMC, and Gf. The isolated components of WM could not uniquely explain the WMC-Gf relationship. Instead, the common variance shared across these components underpins the relation between WMC and Gf.
Chapter 3 examined diverse conceptualizations of binding via structural equation modelling. Item-to-item and item-to-location binding were best represented by a single binding maintenance latent variable. This loaded onto a higher-order binding factor that was statistically indistinguishable from Gf, highlighting the importance of binding maintenance and revealing process-general binding is isomorphic with Gf.
Chapter 4 investigated cognitive dedifferentiation in a cross-sectional sample (ages 18–85). Age-weighted models estimated non-linear partial correlations between WMC, Gf, and processing speed components. The partial correlation between WMC and Gf increased to near unity in older adulthood regardless of processing speed, suggesting speed is not solely responsible for cognitive dedifferentiation.
Ultimately, no single ability could completely account for the WM-Gf relationship. Binding maintenance captured a substantial proportion of their shared variance. These findings challenge single-mechanism theories of fluid cognition and highlight the utility of network models in mapping cognitive relationships.