Stimulating prefrontal cortex facilitates training transfer by increasing representational overlap

Abstract

A recent hypothesis characterizes difficulties in multitasking as being the price humans pay for our ability to generalize learning across tasks. The mitigation of these costs through training has been associated with reduced overlap of constituent task representations within frontal, parietal, and subcortical regions. Transcranial direct current stimulation, which can modulate functional brain activity, has shown promise in generalizing performance gains when combined with multitasking training. However, the relationship between combined transcranial direct current stimulation and training protocols with task-associated representational overlap in the brain remains unexplored. Here, we paired prefrontal cortex transcranial direct current stimulation with multitasking training in 178 individuals and collected functional magnetic resonance imaging data pre- and post-training. We found that 1 mA transcranial direct current stimulation applied to the prefrontal cortex paired with multitasking training enhanced training transfer to spatial attention, as assessed via a visual search task. Using machine learning to assess the overlap of neural activity related to the training task in task-relevant brain regions, we found that visual search gains were predicted by changes in classification accuracy in frontal, parietal, and cerebellar regions for participants that received left prefrontal cortex stimulation. These findings demonstrate that prefrontal cortex transcranial direct current stimulation may interact with training-related changes to task representations, facilitating the generalization of learning.

Keywords: cognitive training; learning; machine learning; tDCS; transfer.

Fig. 1

Summary of experimental protocol. Pre-training behavioral and neuroimaging sessions were followed by 4 d of combined tDCS and cognitive training according to group allocation. Four groups trained on multitasking and received either sham stimulation, 1 mA left hemisphere PFC stimulation, 1 mA right PFC stimulation, or 2 mA left PFC stimulation. Another group trained on a control training task, RSVP while receiving 1 mA left PFC stimulation. Immediately following these combined stimulation and training sessions were post-training neuroimaging and behavioral sessions. Finally, 1 month later, there was another behavioral assessment. For the full protocol including other scans and behavioral assessments beyond the scope of the current study, see Wards et al. (2023).

Fig. 2

Prefrontal tDCS results in enhanced visual search performance. Changes in visual search by group and set size (as previously reported in Wards et al. [2023]). All 5 groups show similar changes in RTs for set size 8 trials. However, both 1 mA left hemisphere (1 mA LH) and right hemisphere (1 mA RH) tDCS groups that trained on multitasking have greater performance improvements than sham for set sizes 12 and 16. Each point represents an individual’s change in RT from pre–post training. Positive values reflect an improvement in performance from pre–post training. Δ = change in, ** = strong evidence, *** = decisive evidence (Kass and Raferty 1995) for a difference between groups.

Fig. 3

Decreased decoding accuracy for the trained task is associated with improvements on untrained visual search performance. A) ROIs that were conjointly activated by the trained task subcomponents, and that showed significantly greater correlation coefficients between changes in decoding accuracy and visual search performance changes than sham. These regions were the right orbitofrontal cortex (R OFC), cerebellum vermis VI (vermis VI), and 2 regions in the left superior parietal lobe (L SupPL and L SupPL2). B) For the group that received left PFC anodal tDCS, the R OFC changes in decoding accuracy were negatively correlated with the changes in RT on visual search trials with 12 distractors (SS12) and this correlation was significantly different to sham. This same pattern of results with negative correlations between changes in decoding accuracy and changes in set size 12 RTs was observed for the vermis C) VI, D) L SupPL, and E) L SupPL2. The L SupPL2 region also had a negative correlation between changes in decoding accuracy and changes in set size 16 RTs. F) Shading around the regression lines displays 95% confidence intervals. SS12 = set size 12, SS16 = set size 16, L = left, R = right, r = Pearson’s correlation coefficient, n = sample size, Δ = change in.