How anodal transcranial direct current stimulation (atDCS) and cathodal tDCS (ctDCS) affect brain networks is still unclear. Previous fMRI studies have yielded controversial results regarding the effects of atDCS and ctDCS on fMRI activation. The present study hypothesizes that the choice of fMRI paradigm may be a contributing factor to this divergence. Therefore, the present study employed two distinct fMRI paradigms, characterized by varying degrees of complexity: finger tapping as a simple fMRI paradigm and an implicit serial reaction time task (SRTT) as a more challenging paradigm. Seventy-five healthy subjects were randomized to receive either atDCS, ctDCS, or sham stimulation during fMRI. The main effects of the blood oxygenation level-dependent (BOLD) signal were contrasted between groups. SRTT, but not FT, was capable of eliciting differences in modulatory effects on the network between groups. Analysis of functional connectivity between ROIs showed that atDCS and ctDCS shared common and distinct SRTT networks. Correlations between BOLD signal (in ROIs) and the reaction time (RT) recorded during fMRI showed that in the atDCS group, faster RT was associated with higher BOLD signal in the most ROIs, while in the ctDCS group, faster RT was mostly associated with lower BOLD signal activity. The sham group exhibited a combination of these associations. We suggest that atDCS accelerates RT by "pushing" the network, while the network response under ctDCS was a "compensatory" response. The polarity of tDCS differentially modulated the adaptive plasticity of remotely connected regions, based on the concept of functional organization of distributed segregated networks.