bioRxiv2023
Aging is accompanied by a decline of multiple cognitive capacities, including working memory: the ability to maintain information online for the flexible control of behavior. Working memory involves stimulus-selective neural activity, persisting after stimulus presentation in widely distributed cortical areas. Here, we unraveled the mechanisms of working memory in healthy older adults and patients with mild cognitive impairment (MCI), a condition associated with increased risk of developing dementia. We studied a sample of 19 older adults diagnosed with MCI and 20 older healthy controls using a combination of model-based behavioral psychophysics, neuropsychological assessment, and magnetoencephalographic (MEG) recordings of brain activity. Twenty-one younger healthy adults were studied with model-based behavioral psychophysics only. All subjects performed a visuo-spatial delayed-match-to-sample working memory task under systematic manipulation of the temporal delay and the spatial distance between successively presented sample and test stimuli. We developed a computational model of the latent dynamics underlying task behavior and fit this to individual behavior. In the model, working memory representations diffused over time, a threshold was applied to produce a match/non-match decision about sample and test locations, and occasional lapses produced random decisions. For the older participants, we related the individual model parameters to a summary measure of individual cognitive integrity obtained from a large neuropsychological test battery, as well as to cortical MEG activity during the delay interval of the working memory task. For all groups, task accuracy decreased with delay duration and sample-test distance. When sample/test distances were small, older adults exhibited larger false alarm rates than younger adults. The behavioral effects were well captured by the model, which explained the age-related differences in terms of a deterioration of the quality of working memory representations, rather than differences in task strategy (i.e., threshold parameter). Task accuracy as well as the parameters governing behavioral stochasticity (diffusion noise and lapse rate combined) were correlated with overall cognitive integrity in the MCI group, but not in the older healthy controls. Individual task accuracy and stochasticity parameters (diffusion noise and lapse rate) were also correlated with stimulus-selective cortical activity during the delay interval, as assessed by decoding of the MEG signals, corroborating their validity as markers of cortical working memory mechanisms. Our findings provide insight into the mechanistic basis of aging-related changes in working memory maintenance and reveal a link between individual working memory dynamics and cognitive integrity in MCI.