Brain activations elicited during task-switching generalize beyond the task: A partial least squares correlation approach to combine fMRI signals and cognition

Paulina Skolasinska,Shuo Qin,Michelle Voss,Hyun Kyu Lee,Chandramallika Basak

DOI: https://doi.org/10.1002/hbm.26804

2024-08-01

Abstract:An underlying hypothesis for broad transfer from cognitive training is that the regional brain signals engaged during the training task are related to the transfer tasks. However, it is unclear whether the brain activations elicited from a specific cognitive task can generalize to performance of other tasks, esp. in normal aging where cognitive training holds much promise. In this large dual-site functional magnetic resonance imaging (fMRI) study, we aimed to characterize the neurobehavioral correlates of task-switching in normal aging and examine whether the task-switching-related fMRI-blood-oxygen-level-dependent (BOLD) signals, engaged during varieties of cognitive control, generalize to other tasks of executive control and general cognition. We therefore used a hybrid blocked and event-related fMRI task-switching paradigm to investigate brain regions associated with multiple types of cognitive control on 129 non-demented older adults (65-85 years). This large dataset provided a unique opportunity for a data-driven partial least squares-correlation approach to investigate the generalizability of multiple fMRI-BOLD signals associated with task-switching costs to other tasks of executive control, general cognition, and demographic characteristics. While some fMRI signals generalized beyond the scanned task, others did not. Results indicate right middle frontal brain activation as detrimental to task-switching performance, whereas inferior frontal and caudate activations were related to faster processing speed during the fMRI task-switching, but activations of these regions did not predict performance on other tasks of executive control or general cognition. However, BOLD signals from the right lateral occipital cortex engaged during the fMRI task positively predicted performance on a working memory updating task, and BOLD signals from the left post-central gyrus that were disengaged during the fMRI task were related to slower processing speed in the task as well as to lower general cognition. Together, these results suggest generalizability of these BOLD signals beyond the scanned task. The findings also provided evidence for the general slowing hypothesis of aging as most variance in the data were explained by low processing speed and global low BOLD signal in older age. As processing speed shared variance with task-switching and other executive control tasks, it might be a possible basis of generalizability between these tasks. Additional results support the dedifferentiation hypothesis of brain aging, as right middle frontal activations predicted poorer task-switching performance. Overall, we observed that the BOLD signals related to the fMRI task not only generalize to the performance of other executive control tasks, but unique brain predictors of out-of-scanner performance can be identified.