The dynamic interplay between overt attention and visual memory in the semantic representation of real-world objects across cognitive ageing

Every day, we interact with countless objects embedded within rich and dynamic environments. To navigate this complexity, the visual system organizes incoming information into structured representations, connecting objects with each other and with the broader contexts in which they appear. This doctoral thesis examines how visual attention and memory interact in naturalistic settings to support the construction of such representations, and how these processes are affected by cognitive ageing. I first focused on the associative organization of semantic memory, exploring how conceptual relationships between real-world objects are structured along categorical and functional dimensions, reflecting shared features within a taxonomic category or co-occurrence in everyday, action-based contexts (Chapter 3). Although both dimensions are central to conceptual organization, it remains unclear whether they constitute distinct representational systems or lie along a continuum of semantic relatedness, where one emerges as dominant and more stable across ageing. Using a novel adaptation of the Visual Paired Comparison paradigm, two eye-tracking experiments revealed that semantic associations are implicitly encoded and later reactivated during visual exploration, leaving measurable traces in gaze behaviour that reflect the accessibility and robustness of their underlying representations. Functional relations exerted a stronger influence on memory-guided attentional dynamics, particularly in older adults, where they appeared more resilient than categorical ones, due to their grounding in everyday experience and action-based affordances, which likely promote the integration of objects into holistic, event-based representations. These findings suggest an implicit hierarchy in semantic knowledge, where patterns of functional co-occurrence form the foundation for broader categorical generalizations. Then, I examined object-scene semantic relationships to understand how objects are contextually integrated within the broader unit of a scene (Chapter 4). A central question concerns whether, and how rapidly, high-level semantic information can be accessed from the visual periphery, and how such information supports memory formation. Across three change-detection experiments, semantically inconsistent objects (e.g., a hairdryer in a kitchen) attracted gaze earlier than consistent ones (e.g., a toaster) and predicted higher detection accuracy. These results suggest that the visual system exploits the statistical regularities of real-world environments to guide overt attention and construct short-term representations during naturalistic viewing. Crucially, these effects emerged across younger adults, healthy older individuals, and participants with Mild Cognitive Impairment (MCI), indicating that predictive mechanisms of object-scene integration remain robust despite both healthy and pathological ageing. Combining eye-tracking with electrophysiology, I further showed that contextual violations elicit brain responses time-locked to the fixation preceding object foveation (N400 component), an effect observed in both younger and older adults. These results challenge models proposing that extrafoveal guidance depends solely on low-level visual features or that ageing reduces peripheral processing, revealing instead that high-level semantic analysis operates pre-attentively across the visual field and remains functionally preserved throughout the lifespan. Beyond their theoretical contributions, these findings highlight the importance of using naturalistic paradigms that capture the richness of real-world vision and can reveal mechanisms that are spared rather than impaired in cognitive ageing. In line with this perspective, I developed and validated two databases of visual stimuli, including object pairs (PAIRS) and real-world scenes (VISIONS), carefully normed along perceptual and conceptual dimensions to balance control and ecological validity in my empirical investigations (Chapter 2). Overall, this thesis advances a unified account of visual cognition as a meaning-driven and predictive system. Integrating behavioural, oculomotor, and electrophysiological evidence, it demonstrates that the semantic guidance of attention and memory remains remarkably robust with age, underscoring its role as a core mechanism in the architecture of the human mind.