Cognitive Reserve, according to a recent consensus definition of the NIH-funded Reserve and Resilience collaboratory (https://reserveandresilience.com/), is constituted by any mechanism contributing to cognitive performance beyond, or interacting with, brain structure in the widest sense. To identity multivariate activation patterns fulfilling this postulate, we investigated a verbal Sternberg fMRI task and imaged 181 people with age coverage in the ranges 20-30 (44 participants) and 55-70 (137 participants). Beyond task performance, participants were characterized in terms of demographics, and neuropsychological assessments of vocabulary, episodic memory, perceptual speed, and abstract fluid reasoning. Participants studied an array of either 1, 3, or 6 upper-case letters for 3 seconds (=encoding phase), then a blank fixation screen was presented for 7 seconds (=maintenance phase), to be probed with a lower-case letter to which they responded with a differential button press whether the letter was part of the studied array or not (=retrieval phase). We focused on identifying maintenance-related activation patterns showing memory-load increases in pattern score on an individual-participant level for both age groups. We found such a pattern that increased with memory load for all but one person in the young participants (p<0.001), and such a pattern for all participants in the older group (p<0.001). Both patterns showed broad topographic similarities; however, relationships to task performance and neuropsychological characteristics were markedly different and point to individual differences in Cognitive Reserve. Beyond the derivation of group-level activation patterns, we also investigated the inter-subject spatial similarity of individual working-memory rehearsal patterns in the older participants’ group as a function of neuropsychological and task performance, education and mean cortical thickness. Higher task accuracy and neuropsychological function was reliably associated with higher inter-subject similarity of individual-level activation patterns in older participants.

The data involves functional fMRI for 181 participants. Pre-processing and analytic details can be found in the manuscript.

Usage notes are included and provided in a README.doc file. They contain:

- an explanation of variables in a Matlab data archive

- a listing of most important Matlab .m files and how to execute them to recreate the main brain-behavioral results of the paper

- a listing of 4 thresholded and sign-weighted voxel-wise Z-images that can be visualized, for instance with the package MRIcron