Dissertations, Theses, and Capstone Projects

Date of Degree

2-2026

Document Type

Doctoral Dissertation

Degree Name

Doctor of Philosophy

Program

Psychology

Advisor

Nancy S. Foldi

Committee Members

Veronica Hinton

Laura Rabin

Carolyn Pytte

Jason Bock

Subject Categories

Clinical Psychology | Cognitive Psychology | Psychology

Keywords

neuropsychology, mild cognitive impairment, serial position effect

Abstract

Verbal episodic memory impairment (i.e., linguistic-mnemonic deterioration) appears as one of the earliest and most sensitive cognitive markers of preclinical Alzheimer’s disease (AD). However, standard scoring methods for wordlist memory tasks may not optimally capture the subtle cognitive processes most predictive of clinical progression. This dissertation holistically assessed whether metrics derived from Hierarchical Bayesian Cognitive Process (HBCP) modeling—latent cognitive variables aligning with features of a wordlist memory test—offer superior prognostic value for identifying individuals at higher risk of clinical decline over 36 months compared to traditional recall scores and serial position effect process metrics. The study used statistical modeling techniques to analyze and compare verbal memory data from two wordlist memory tasks (RAVLT, ADAS-Cog) in Alzheimer’s Disease Neuroimaging Initiative (ADNI) participants spanning cognitively normal, subjective cognitive impairment, and mild cognitive impairment diagnostic categories at baseline. Covariate-adjusted logistic regression models included demographic factors (age, sex, education) and baseline clinical status to improve prediction accuracy, with model performance assessed using various statistical tests and interpreted through a clinical lens. Results showed partial, task-dependent support for HBCP parameters’ cross-task generalizability (Aim 1). In the RAVLT sample (N = 303), all three cognitive parameters showed robust prognostic associations with large effect sizes, while in the ADAS-Cog sample (N = 301), only the latent probability of recall of an episodic memory on the delayed recall task remained significantly prognostic. Pooled generalized estimating equation models found no statistically significant Task × M interactions, which supports measurement invariance after standardization, though mixed findings warranted a nuanced interpretation. In the RAVLT sample (Aim 2.1), hierarchical model comparisons indicated that the HBCP M model had discrimination statistically equivalent to Traditional scoring but was significantly better than serial position effect (SPE) Process scoring. Information criteria favored Traditional over HBCP M, highlighting Traditional’s parsimony advantage, although the Full model had the highest discrimination at the expense of increased complexity. These findings emphasize that the prognostic value of HBCP parameters largely depends on task structural features. Within-region comparison (Aim 2.2) showed that quantitative cognitive process composites offered only slight benefits over SPE regional ratios in the primacy domain, and no benefits in recency. The between-region comparison (Aim 2.3) strongly supported primacy’s superiority over recency, aligning with theories and prior literature indicating the sensitivities of early-list encoding processes to incipient AD. This dissertation merged cognitive neuropsychological testing with various statistical methods to help inform optimal approaches for identifying and characterizing risk in individuals at the preclinical stages of AD. Future research directions include seeking external validation in independent, more demographically diverse cohorts and biomarker correlation studies to establish neurobiological construct validity of HBCP parameters across various clinical contexts.

Download

This work is embargoed and will be available for download on Monday, February 01, 2027

Share

COinS