This randomised, double-blinded, sham-controlled tDCS study highlights the importance of frontal networks in learning a complex dynamic balance task. Our results demonstrate that the influence of c-tDCS over these networks during a long-term motor learning process caused higher performance variability compared to the s-tDCS stimulation group. This increase in behavioural variance indicates that the stimulation causally affected (pre-)frontal brain networks 27,28. Moreover, DBT training with concurrent c-tDCS not only resulted in a ‘near’ transfer effect on postural control, but also in ‘far’ transfer on cognitive flexibility known to rely on the prefrontal networks persisting 24 hours after the end of training.
In this study, tDCS applied during DBT practice was aimed at influencing network nodes implicated in long-term DBT learning. Hence, shifting the focus onto the specific task-relevant activation of networks, down-weighing the low anatomical precision of tDCS28,40. These network nodes were selected based on previous findings showing macro- and microstructural properties of PFC-SMA regions predict future DBT learning10,41 also changing in response to DBT practice11,42,43. Although these studies provided evidence of a brain-behaviour relationship between PFC-SMA networks and balance learning, demonstrated using approaches like statistical mediation analyses 44, the neuroimaging findings remain correlative. However, 35 showed a single session of online c-tDCS over the right PFC-SMA region during training has an acute effect on subsequent DBT performance. Here, we extend these previous findings by causally showing PFC-SMA network involvement in long-term balance learning, manifesting itself through increased performance variability 28.
The true direction of the effect of tDCS on performance may be masked/varied across and with-in participants due to dissimilar amplification in neuronal noise, in such cases, the sheer increase in variance (beyond measurement noise) after tDCS may be considered evidence for a cause–effect relationship28. Such behavioural consequences of tDCS may arise due to individual differences in the recruitment of brain networks during task performance leading to differences in excitability modulation 20,28,45. Along with reported within-session, non-linear effects of c-tDCS46, dissimilarities in tDCS induced modulation of cortical excitability may not necessarily translate into behavioural deviations as drastic as performance inhibition. Lack of DBT performance deterioration can therefore be associated with tDCS being a weak direct current and its behavioural effects meagre; making it possible for networks to capably compensate for weak disturbances during online stimulation by adapting to the electric field over time28. The results of this study demonstrate improved DBT performance for both groups over the 3-week training duration; indicating similar task proficiency at the end of practice. Hence, tDCS may have affected the process of learning a complex task rather than altogether changing the learning trajectory.
The prefrontal networks involved in the strategy building aspect of motor learning were the prime target of c-tDCS in our study 24,40. Consequently, participants were not instructed on the most optimal task execution strategy (contrary to a ‘classical’ motor skill learning/training), instead, encouraged to learn the task by discovering their own strategies via trial and error47. Previous studies investigating the mechanisms involved in adopting specific courses of action during learning have associated the anterior PFC in exploration of new possibilities. Here, future outcomes are said to be predicted by tracking alternative options and exploratory switching between courses of actions through extrapolation of short-term trends 7,9. Hence, task complexity and uncertainty of outcomes may dictate the extent of PFC involvement, where selection of appropriate strategies and guiding cognitive resources to implement these strategies is done by integrating and comparing various sequential outcomes 6,9. Owing to the task complexity and the available solution space, the DBT fulfils criteria’s particularly conducive for cognitive processes involved in reinforcement learning, in particular, exploration of solutions achieved through various coordinative whole-body movements. Therefore, we speculate that PFC-dependent networks responsible for exploration of new performance strategies (in the context of learning) were modulated by c-tDCS. This modulation was behaviourally expressed as increased performance variability.
It is suggested that extending learning gains to other untrained tasks is possible only if a shared commonality exists between these tasks, viz., abilities required in executing both tasks, neural processing mechanisms and brain regions 16,17,48. These transfer effects are also theorised to be tied to early phases of structural plasticity within overlapping networks19. The ‘neural overlap hypothesis’ has been supported by evidence from concurrent tDCS during cognitive training resulting in microstructural brain alterations alongside near-transfer behavioural effects 34,49. Since the motor learning paradigm used in this study is capable of inducing structural grey and white matter changes in PFC and SMA regions 11,43,50,51, we further hypothesized it to potentially lead to cognitive transfer effects. Consistent with this hypothesis, we found higher improvement in executive functioning performance (i.e., ΔTMT and TMT-B)52 as a result of DBT training with concurrent rPFC c-tDCS compared to s-tDCS. Both, aerobic exercise on its own53 and a-tDCS over left DLPFC during coordinative exercise 54 have shown a tendency towards TMT performance improvements. Similarly, cognitive training combined with tDCS at an intensity of 1.0-mA augmented both decision-making performance and cognitive transfer55.
Despite a global network involvement in TMT execution56, our regions of interest were restricted to the overlapping PFC-SMA networks involved in DBT learning. We hypothesize the combination of DBT training-induced plasticity, discovery-learning based motor training and tDCS to encourage a rapid network reorganisation and compensation 57–59. This functional compensation probably constituted conditioning new or otherwise inactive networks within the overlapping brain regions leading to an advantageous effect of intervention, absent in the s-tDCS group 60,61. Benefiting from richly connected brain networks supporting a multitude of cognitive functions required in TMT-B execution may have improved the potential for transfer via compensatory mechanisms in the overlapping networks 56,62–64. A combination of brain imaging and stimulation techniques is required to prove the specific functional and structural correlates of PFC involvement in learning and associated transfer.
Contrary to executive functioning, we did not find significant differences between either groups on memory and attention abilities, although positive effects of physical exercise (e.g., coordinative and aerobic exercise) on visuospatial attention, working memory 65, associative memory, spatial cognition 14,15 and visuospatial memory 66 have been observed in previous studies. Note, however, that our results indicate marginally better performance in the attention task (d2-R) exhibited by the c-tDCS group compared to the s-tDCS group. Although this difference did not reach statistical significance. On the other hand, the s-tDCS group showed a tendency towards higher improvements in an SMA-dependent selective interference resolution task (Eriksen flanker task- accuracy interference) as compared to the c-tDCS group, this trend was accompanied by a medium sized effect (Results 2.3.4).
Finally, the observed transfer effects on PFC-SMA-dependent cognitive tasks can be assumed to be due to a shared commonality with the trained task (neural overlap hypothesis)19,48, which changed as a function of the intervention, demonstrating a potential common neural substrate underlying the trained balance task and the transfer task67. This complex motor training engaging higher-order processes may have enabled cognitive improvements by transferring learning gains to untrained tasks. In turn benefiting abilities like information processing, goal-dependent inhibition/ maintenance of responses, formulating strategies based on feedback, distributing attention over multiple strategies, switching between strategies (cognitive flexibility), etc 16,17,48. Findings from 14 demonstrate balance training-induced improvement in memory and spatial cognition attributed to a training that encompassed proprioceptive, visual and motor-based learning. Likewise, a month of slackline training improved vestibular-dependent spatial orientation performance 13 suggesting a positive effect on vestibulo- hippocampal spatial orientation.
Lastly, we also observed a statistical tendency towards larger near-transfer effects to an untrained balance task (Nintendo Wii header game- advanced level) in the c-tDCS group compared to the s-tDCS group. Interestingly, consistent with the ‘neural overlap hypothesis’, in the c-tDCS but not in the s-tDCS group we observed a medium-sized positive correlation between DBT performance variability and Wii scores. Such near motor transfer effects have recently been observed by 68, manifested as improved cross-limb transfer from the trained to the untrained hand after anodal tDCS over rM1 in older adults. Similarly, we hypothesize that participants in our study were able to successfully use the movement solutions learned during DBT training onto an untrained balance task which also requires a comparable movement pattern in terms of body’s centre of mass (COM) control and displacement. 69–71 emphasize introduction of variation during practice as a key aspect in eliciting new movement solutions enabling a degree of transfer beyond the practiced solutions. However, further studies are required to support the role of movement variability to improve transfer during stabilometer learning.
Although the results of this study highlight the importance of the frontal networks in learning a complex task, we are unable to disentangle the contributions of PFC from those of SMA as both these regions have been implicated with undergoing learning-induced structural changes. Our cognitive transfer results do point towards higher PFC involvement but we were not able to definitively outline the specific contributions of these regions. Further work utilizing a combination of tDCS and neuroimaging may aid in explicitly mapping stimulation-induced changes at the neuronal and network levels. Linking these brain changes to the behavioural effects would be the natural subsequent step in order to unravel the complexity of the underlying brain-behaviour relationship. Stimulating an alternative brain region is advised in order to ascertain that the observed effects emanate solely as a result of interference within the regions of interest 28,29. However, this control condition was not included since we intended on influencing the networks previously implicated in learning the complex DBT. In light of the recently revealed predispositions to improved learning abilities 10,41, heterogeneity of participants in the form of genetic makeup, brain structure and environmental diversity requires consideration 72. The solitary effect of tDCS on cognitive abilities without the influence of training is an aspect that could help differentiate between the cumulative effect of tDCS and training observed in this study.
Our results provide new evidence for PFC-SMA involvement during long-term DBT practice. Specifically, we show that interfering with these networks using c-tDCS leads to increased performance variability, potentially indicating a causal involvement of PFC-SMA networks in DBT learning28. Against the background of ‘neural overlap hypothesis’, we interpret the observed tDCS-effects on motor and cognitive performance as tDCS effects pertaining not only to the trained tasks, but also to the untrained tasks which rely on overlapping brain networks. The conclusions drawn through this study reinforce the positive impact of physical activity on cognition through the synergistic neural networks sub-serving both motor processing and cognitive functioning. An understanding of this brain-behaviour relationship may prove valuable not only in promoting overall health through exercise but also support healthy aging by means of mobilizing neural resources to remedy dysfunction.