The present study tested the effects of tDCS-induced cortical excitability changes (anodal, cathodal and sham tDCS) in DLPFC and M1 on two different set shifting tasks (cognitive and motor). For the cognitive task, anodal stimulation was found to increase performance as indicated by an RT decrease compared to both sham and cathodal conditions. Although cathodal tDCS decreased overall performance, there was no statistically significant difference when compared to sham tDCS. However, in the motor task, cathodal stimulation significantly decreased performance when compared to sham and anodal stimulation.
One important finding is that these results are independent of the stimulation site, suggesting a non-specific site effect probably due to interactions among the several neural networks that have been shown to be activated when performing set shifting/task switching tasks 
. Task switching research has demonstrated that performing one task and then another could activate a common frontal parietal network 
. Moreover, both motor and executive functioning areas could be responsible for distinct cognitive processes involved in a broader cognitive control process 
. However, that does not entirely explain the tDCS effects found in this study, especially the task polarity interaction.
One hypothesis is that the effects are dependent on the level of activation of this network. In other words, for the cognitive task, in which the demand on motor systems is less intense, anodal tDCS was able to enhance performance as the system was likely engaging a more reduced neural network as compared to that engaged by the motor task. In fact, for the motor task, because the co-activation of motor and executive areas was likely more intense, an increase in activity induced in only one area was not sufficient to enhance performance. On the other hand, the cathodal-induced excitability decrease in motor or prefrontal areas was associated with a performance reduction in the motor task due to activity reduction in one region of this highly engaged network required for performance of both tasks.
There are also alternative hypotheses to explain our results. For instance, the lack of specific effects might be explained by the lack of focality of the tDCS. In this scenario, DLPFC tDCS induced similar effects as M1 tDCS due to the lack of focality. However, modeling and behavioral studies tend not to support this alternative explanation, as they show that the peak of the current is induced under the electrode 
and also that DLPFC and M1 tDCS induce different behavioral effects 
. Alternatively, as the “reference” electrode was positioned over the contralateral supraorbital area, it is also possible that this electrode exerted an effect on our results. This hypothesis arises because Brodmann Area (BA) 10 has been associated with these particular types of tasks 
and because tDCS studies have shown effects on cognitive processing induced by that particular site 
. Future studies need to assess other electrode montages to rule out this effect, namely, by using extra-cephalic reference electrodes. Using smaller electrodes will also be a future option for testing non-specific results.
In terms of the filtering of irrelevant information, the pattern found in this study was Alone<Neutral<Incongruent, which is consistent with previous studies 
. The explanation that has been provided for this is that as the distracting set gets more challenging, there is an increased demand on filtering 
. This study shows that anodal and cathodal tDCS both modulate cognitive and motor tasks. They had a consistent effect on results independent of the site of stimulation (anodal improved and cathodal decreased task performance), suggesting that the cortical stimulation is modulating this highly engaged network involved in set shifting.
There were no specific effects of tDCS on the filtering of irrelevant information or on Shift costs. There were also no errors related to tDCS. The error effects found were related to shift or filtering conditions, and are being interpreted as more demand on resources due to normal task performance.
Future research using an fMRI paradigm should explore the assumption that cortical tDCS could interfere with shift ability by affecting this highly engaged network (with cortical and eventually subcortical processing) to establish the cortical tDCS effects and possible cortical-striatum interactions. In addition, future studies should also explore neuromodulation of cortical-subcortical activity in different pathologies with set shifting impairments, namely Obsessive-compulsive disorder, eating disorders, Parkinson's and Huntington's disease, as well as in aging.
Future research should also focus on the effects of tDCS on dopamine receptors using these set shifting tasks, as the administration of D2 antagonists in healthy subjects 
showed an effect on set shifting similar to the one found in this study with cathodal stimulation.
One of the limitations of this study was the lack of statistical power to include all the factors in a full multifactorial analysis. Thus some of the results need to be seen as exploratory and need to be confirmed with larger sample sizes. Also future research should apply tDCS during the actual task, in order to compare the results from learning phase to actual performance, as there could be specific learning phase effects 
.Also, the cognitive task took longer to perform than the motor one. This time difference found in performance between tasks may be a limitation of the present study. Future studies should match the duration of the cognitive and the motor task (possibly by establishing a time limit rather than number of trials). In conclusion, the present study found that both anodal and cathodal tDCS can modulate a cognitive–motor task. The non-specific site effects could be related to an interaction within this neural network, to the network demand involved in these two tasks, or to the enrollment of the right supraorbital in this highly engaged network. Finally, a single session of tDCS to the left DLPFC or to M1 (or the right supraorbital) seemed to have a greater result on the speed of changing sets than on Shift costs, either by reducing the number of errors or by increasing the efficacy of irrelevant set filtering.