Using ~200 retired people above 65 from the CamCAN cohort of healthy ageing, we previously showed that mid-life activities outside the workplace make a unique contribution to late-life cognition (fluid intelligence), over and above education and current activities (D. Chan et al, 2018; https://doi.org/10.1016/j.neurobiolaging.2018.06.012; https://osf.io/32gme/). In that paper, we also showed that higher levels of mid-life activities reduce the relationship between total grey-matter volume (TGM) and cognition in late-life, i.e, function as a form of "cognitive reserve". In this new project, we hope to identify a functional brain correlate of cognitive reserve in the same sample. Based on work by (a different) M. Chan and colleagues (https://doi.org/10.1073/pnas.1415122111, https://doi.org/10.1073/pnas.1714021115), we hypothesise that mid-life activities affect the functional segregation (FS) of resting-state networks (in late-life). We will start with resting-state fMRI data from CamCAN, and use methods similar to M. Chan et al to define FS for each participant, restricted to what M. Chan et al call "associative" networks. The methods for estimating FC are already specified here: https://github.com/ethanknights/cognitiveReserveMidlife-Connectivity. We predict that: 1. Mid-life activities (from the LEQ retrospective questionnaire) will positively predict FS (in multiple linear regression, adjusting for age and sex). 2. This positive relationship will remain even when adjusting for the other 5 LEQ scores (one of which includes education) 3. FS will predict cognitive health (fluid intelligence from Cattell test, as in our original D. Chan et al paper), adjusted for age and sex 4. If 1+3 are true, then FS should mediate the effect of mid-life activities on cognitive health 5. If 4 is true, then FS will also moderate the relationship between Total Grey Matter (as in D. Chan et al) and cognitive health (reducing slope when FS is high). We will report p-values as well as Bayes Factors for the null, using the same R code as in "Paper_Analysis_Revised.R" on https://osf.io/q9bj4/, with additional use of "regressionBF" in R "BayesFactor" package to estimate BF01. Given that fMRI only detects slow changes in brain activity, we will repeat tests 1-5 with an alternative measure of FS from CamCAN's resting-state MEG data. The estimation of FS for MEG is already detailed here: https://github.com/ethanknights/connectivity-multimodal/tree/main/rest/MEG. Finally, given that the above two definitions relate to "static" (or average) connectivity over ~9mins of rest, we will also conduct an exploratory Canonical Correlation Analysis (CCA) relating mid-life activities to "dynamic" measures of MEG connectivity based on our previous work (Tibon et al, 2021; https://doi.org/10.1016/j.neurobiolaging.2021.01.035). We will use CCA to identify a profile of dynamic metrics that relate to mid-life activities, adjusted for age and sex, and then repeat tests 3-5 above.