Compared to young adults, older adults consistently demonstrate reduced episodic memory performance, that is, poorer memory for unique events (Tulving, 1983). As we discuss below, age-related episodic memory decline has been linked to age differences in the integrity of the brain’s white matter (i.e., structural connectivity), differences that likely impact the efficiency of communication between different neural regions.

Contemporary magnetic resonance imaging (MRI) studies of white matter microstructure typically employ diffusion tensor imaging (DTI) methods. These methods allow non-invasive investigation of inter-regional structural connectivity (for reviews, see Sporns, 2013; Yeh et al., 2020), providing estimates of microstructural integrity by measuring water diffusivity (Assaf and Pasternak, 2008). The most commonly employed DTI metric - fractional anisotropy (FA) – is a measure of the relative rate of water diffusion along as opposed to across the longitudinal axis of an axon. Relatively higher FA is widely assumed to indicate relatively greater white matter integrity (Madden et al., 2009).

Although widely employed, conventional DTI metrics are susceptible to contamination by partial volume effects caused by the juxtaposition of extracellular ‘free water’ and white matter. DTI metrics are more likely to be biased for white matter tracts adjacent to ventricles, and such effects are likely to be exaggerated in older adults because of age-related white matter atrophy and a corresponding partial volume effect (Alexander et al., 2007, Giorgio et al., 2010, Vos et al., 2011). Of importance, partial volume effects can lead to an underestimation of FA (Metzler-Baddeley et al., 2012; Pfefferbaum et al., 2003) and an overestimation of the effects of age on white matter integrity (Chad et al., 2018). It is, however, possible to correct for the contribution of extracellular free water to DTI metrics (Pasternak et al., 2009). Studies employing this correction procedure have reported that, compared with uncorrected FA estimates, the procedure increases estimates of FA both in individual tracts and across the whole brain (Albi et al., 2017, Bergamino et al., 2017, Edde et al., 2020). Moreover, Chad et al. (2018) reported that the relationship between age and whole brain mean FA was substantially reduced after free water correction (r = - 0.51 vs. r = - 0.25 before and after correction respectively) in a group of 212 participants aged between 39 and 92 years old. In light of the evidence reviewed above, we applied a free water correction procedure to the DTI data acquired in the current study.

Here, we focus on the fornix and parahippocampal cingulum, two limbic tracts that connect the hippocampus - long known to play a crucial role in episodic memory (Eichenbaum, 2017) - with a diverse range of other brain regions (Bubb et al., 2018, Mark et al., 1995, Schmahmann and Pandya, 2009). The fornix is the major tract connecting the hippocampus with a number of sub-cortical structures, including the septum and, via the mammillary bodies, the anterior thalamus (Amaral and Lavenex, 2007; Senova et al., 2020). By contrast, the parahippocampal cingulum (hereafter, cingulum), which forms part of the cingulum bundle, has widespread connections with frontal, parietal, occipital and temporal cortical regions (Jones et al., 2013, Maldonado et al., 2020). Damage to either tract has been reported to lead to memory impairment, albeit less consistently in the case of the cingulum (for reviews, see Bubb et al., 2018; Douet and Change, 2015; Benear et al., 2020).

The great majority of studies investigating associations between memory performance and the integrity of the fornix and cingulum have employed conventional (i.e. uncorrected for free water) DTI metrics. In middle-aged and older adults, higher fornix FA has been reported to predict better performance on tests of face recognition (Ly et al., 2016), verbal recall (Hartopp et al., 2019, Metzler-Baddeley et al., 2011), visual recognition (Hartopp et al., 2019), visual recall (Metzler-Baddeley et al., 2011) and autobiographical memory (Memel et al., 2020). Higher fornix FA has also been linked to lower rates of false recollection in older adults (Chamberlain et al., 2021). Several studies have examined the relationship between fornix integrity and memory across the adult lifespan (Alm et al., 2020, Bennett et al., 2015, Bennett and Stark, 2016, Sasson et al., 2013, Henson et al., 2016). In four of these studies, no relationship between fornix integrity and memory performance was evident after controlling for age (Ly et al., 2016, Metzler-Baddeley et al., 2011, Bennett et al., 2015, Henson et al., 2016). By contrast, a positive correlation between fornix FA and measures of autobiographical memory and visual recognition was reported by Memel et al. (2020) and Hartopp et al. (2019) respectively.

Findings regarding the relationship between cingulum integrity and memory performance are also inconsistent (for review, see Bubb et al., 2018). Whereas two studies reported a positive relationship (after controlling for age) between cingulum FA and memory performance in middle-aged and older adults (Ezzati et al., 2016, Li et al., 2020), no significant relationship between these variables was reported in three other studies (Bennett et al., 2015, Hartopp et al., 2019, Memel et al., 2020).

Unlike the cross-sectional studies reviewed above, studies employing longitudinal designs allow identification of relationships between neural metrics and within-person, age-related changes in cognitive performance. Hence, such studies provide an opportunity to establish whether a given metric is a predictor of age-related cognitive decline. To date, however, only a handful of studies have examined whether fornix or cingulum integrity is predictive of longitudinal memory change in cognitively healthy older adults (Lancaster et al., 2016; Rabin et al., 2019a, b; Song et al., 2018). In the case of the fornix, Rabin et al. (2019b) reported that higher FA at baseline was associated with a smaller decline in verbal memory recall over 4 years, whereas no such relationship was identified in Lancaster et al. (2016). Using radial diffusivity as the DTI metric, Song et al. (2018) reported a nonsignificant association between fornix integrity and 4-year memory change. In a similar vein, of the two studies that examined the cingulum (Lancaster et al., 2016; Rabin et al., 2019a), only Lancaster et al. (2016) reported a significant relationship between cingulum FA and memory change over a 3-year follow-up period.

To our knowledge, only two studies have examined associations between free water-corrected DTI metrics derived from the fornix and cingulum and either memory performance or memory change in older adults (Ji et al., 2019; Archer et al., 2020). Ji et al. (2019) reported a significant correlation between water-corrected fornix FA and memory performance in patients with Alzheimer’s Disease (AD) or amnesic mild cognitive impairment (aMCI). By contrast, Archer et al (2020) reported that neither fornix or cingulum water-corrected FA metrics were associated with memory performance in a sample comprising a mixture of cognitively healthy older adults and older adults with aMCI. However, water-corrected cingulum FA, but not fornix FA, was predictive of memory decline over 5 years. Given that both of these studies included cognitively diverse samples, it is unclear whether these DTI metrics are sensitive to memory performance or memory change in older samples comprising only cognitively healthy participants.

In the present study, we examined associations between free water-corrected FA (henceforth, FA) derived from the fornix and cingulum, memory performance, and 3-year longitudinal memory change in a sample of cognitively healthy older adults. Given that a tract-wise relationship between FA and memory might merely reflect a more general association between whole brain white matter integrity and cognition (Bennett and Madden, 2014; Rabin et al., 2019a), we also evaluated the specificity of the relationships between the two FA metrics and longitudinal memory change after controlling for FA in other white matter tracts as well as measures of performance in other cognitive domains.