Nonlinear Lifespan Trajectories of Memory and Fluid Reasoning: A Longitudinal GAMM Study

Longitudinal studies in cognitive aging provide critical advantages over cross-sectional designs, with growing consensus that flexible nonlinear approaches are needed to capture the complexity and heterogeneity of lifespan neural change. In the present study, we leveraged three waves of fMRI data and implemented generalized additive mixed models (GAMMs) to characterize linear and nonlinear trajectories of brain activation across the cognitive domains of memory and fluid reasoning, and to identify regions displaying age-varying cognition–brain coupling. Participants (memory: N = 431; fluid reasoning: N = 441; ages 20–80+) from the Reference Ability Neural Network study completed up to three fMRI assessments over 0–12 years. Domain-level activation maps were parcellated using the Glasser atlas and modeled with tensor-product smooths of baseline age and time, alongside age-varying coefficient terms linking within-person activation fluctuations to in-task performance. Cognition-only models showed robust, approximately monotonic age-related declines in both domains, with no reliable Age × Time interaction. In contrast, activation models revealed pronounced nonlinear Age × Time effects concentrated almost exclusively in posterior association cortex. A particularly notable example was the right POS2 (precuneus/cuneus), which exhibited a midlife reduction in activation, consistent with evidence implicating the precuneus as an early site of functional vulnerability in aging. Beyond these activation trajectories, cognition–activation coupling analyses uncovered age-varying relationships across posterior midline and dorsal parietal regions as well as left rostral Area 6 and IFJp, including several age-dependent changes in the strength and sign of coupling. Posterior midline regions shifted from positive to weaker or negative coupling with age, whereas left rostral Area 6 and IFJp showed the inverse pattern, with negative coupling in youth that diminished or reversed in older adulthood. Together, these findings highlight posterior association cortex as a major site of functional reorganization across the adult lifespan and demonstrate that flexible GAMM approaches reveal developmental inflection points and shifts in brain–behavior coupling that remain undetectable using linear or cross-sectional methods.