Healthy aging is associated with neural dedifferentiation, a reduction in the neural selectivity of higher-level sensory cortex (for reviews, see Koen and Rugg, 2019; Koen et al., 2020). Prior findings indicate that lower neural selectivity is associated with poorer performance on tests of episodic memory and fluid intelligence (e.g., Park et al., 2010; Koen et al., 2019; Srokova et al., 2020, Srokova et al., 2024) and is accompanied by higher levels of cortical tau in cognitively healthy older adults (Maass et al., 2019, Sheng et al., 2024). These findings support the proposal that age-related neural dedifferentiation significantly contributes to age-related cognitive decline. However, the mechanisms underlying dedifferentiation are poorly understood.

When operationalized at the level of stimulus categories, lower neural selectivity is consistently observed in scene-selective cortical regions (Voss et al., 2008, Carp et al., 2011, Zheng et al., 2018; Koen et al., 2019; Srokova et al., 2020, Srokova et al., 2024), whereas null age effects are commonly reported for objects, words, or faces (Chee et al., 2006, Koen et al., 2019, Payer et al., 2006, Srokova et al., 2020, Srokova et al., 2024, Voss et al., 2008, Zheng et al., 2018 but see also Park et al., 2004, Park et al., 2012, Zebrowitz et al., 2016). An important qualification is that these patterns of age differences may differ when neural differentiation is operationalized at different levels of analysis, such that reduced differentiation may be observed not only for scenes but also for faces and objects in cases when differentiation is computed at the level of individual stimuli (e.g., Goh et al., 2010; Lee et al., 2011; Pauley et al., 2024; Srokova et al., 2024). Therefore, neural differentiation at different levels of analysis might depend upon different neural mechanisms (see Discussion for more detail). Here, we focus on differentiation at the category level.

In animal models, neural dedifferentiation has been linked to broadened receptive fields and reduced single neuron selectivity in the visual cortex of aging macaques (Schmolesky et al., 2000, Leventhal et al., 2003, Li et al., 2008). These effects have been associated with age-related declines in GABAergic inhibition, and extending this work, human studies have found that lower cortical GABA levels correlate with reduced neural specificity at the level of visual categories and across large-scale networks (Lalwani et al., 2019, Cassady et al., 2019, Cassady et al., 2020, Chamberlain et al., 2021). In a similar vein, computational models of cognitive aging propose that lower neural selectivity reflects the broadening of neural tuning prompted by age-related alterations in the availability of cortical dopamine (Li et al., 2001). However, because declines in neurotransmitter availability are typically considered to operate cortex wide, this proposal provides little insight into why age-related dedifferentiation, at least at the categorical level, should be apparent in some cortical regions but not others.

As we have proposed previously (Srokova et al., 2020, Srokova et al., 2024), a potential explanation for reduced category-level specificity for scene stimuli relates it to a decline in the ability to perceive and bind features of relatively complex images, processes that could be reflected in eye movement behaviors. Eye movements support the accumulation and integration of visual information as it becomes encoded into memory (e.g., Loftus, 1972; Henderson et al., 2005; Damiano and Walther, 2019). Neuroimaging evidence suggests that more frequent gaze fixations are associated with higher stimulus-evoked fMRI BOLD signals in early and higher-level visual cortex (Henderson and Choi, 2015, Henderson et al., 2020), as well as in the hippocampus (Liu et al., 2017, Liu et al., 2018, Liu et al., 2020). In light of these findings, it is intriguing that older age is associated with marked alterations in eye movement behaviors (for review, see Ryan and Shen, 2020, Wynn et al., 2020). In one relevant study (Liu et al., 2018), younger and older adults underwent simultaneous fMRI with eye-tracking as they viewed face stimuli. The data revealed that older adults made more frequent eye movements (i.e., higher trial-wise fixation counts), and that the relationship between eye movements and fMRI BOLD signals in the hippocampus and fusiform face area was weaker than in younger adults. These findings suggest that gaze fixations become less functionally specific with increasing age, which would contribute to age-related reductions in the ability to bind visual information into a coherent memory representation.

Why would age differences in oculomotor behaviors contribute to reduced category-level neural differentiation for scenes but not objects? A key consideration pertains to the fact that these two categories of visual stimuli engage differential eye gaze scanning behaviors. While objects depict a single item devoid of contextual information, perceiving scenes requires the integration of multiple spatially distributed visual features. Accordingly, while object identity can be inferred from as little as a single fixation directed to the focus of attention, scenes engage geometric and relational processing among multiple features, with much of the contextual information derived from peripheral vision (Voss et al., 2017). In other words, scene perception requires more frequent and more dispersed gaze patterns relative to objects. Altered eye movement patterns may therefore render neural differentiation for scenes more sensitive to aging by compromising the quality and specificity of visual input during encoding.

Strategic and functionally specific fixation behaviors in younger adults supports efficient feature binding. In contrast, less discriminating gaze patterns in older adults may compromise the formation of distinct representations for the stimuli. Indeed, recent evidence suggests that even in the absence of visual input, scene- and face-selective cortical regions can be activated (as measured by fMRI BOLD) by eye movement patterns alone by instructing participants to move their eyes in patterns typically elicited during scene or face viewing (Wang et al., 2019). This raises the possibility that age-related alterations in eye movements may cause gaze patterns to become less ‘scene specific’. Reduced gaze specificity may subsequently compromise the ability to encode the most informative spatial features of the visual scene, leading to reduced specificity of the neural representations arising from the scene-selective cortex.

In the present study, we employed fMRI with simultaneous eye-tracking to test the hypothesis that lower neural selectivity in older adults during memory encoding is associated with age differences in eye movements. Participants studied word-object and word-scene stimulus pairs prior to a memory test. As in our prior work (Koen et al., 2019; Srokova et al., 2020, Srokova et al., 2024), neural selectivity was quantified using multivoxel pattern similarity analysis. We predicted that older adults would exhibit lower category-level neural selectivity for scene but not object stimuli, and this would be accompanied by an age-related increase in gaze fixation count during image viewing. Furthermore, neural selectivity was predicted to demonstrate a positive association with fixation counts in young adults. However, under the assumption that eye movements become less functionally specific in older age, we expected that this association would be weaker in older adults, particularly during scene viewing.

We considered two additional variables in exploratory analyses. Firstly, prior evidence suggests that older adults explore peripheral scene content less extensively than their younger counterparts (e.g. Ball et al., 1988; Romano Bergstrom et al., 2016), potentially limiting the amount of information represented by the scene-selective cortex. We explore this possibility with an analysis of ‘gaze dispersion’ that quantifies the spatial extent of visual sampling across the two age groups. Second, pupil size is thought to be a proxy for arousal-linked enhancement of norepinephrine (NE) availability, which is proposed to support neural gain and attentional tuning (for review, see Mather and Harley, 2016; NE-related mechanisms supporting neural gain may also underlie age-related neural dedifferentiation). Accordingly, we conducted a second exploratory analysis to test for age differences in pupil responses during memory encoding and their potential relationship with neural differentiation.