Adapting one’s decisions to changing contexts is a critical component of flexible behaviour. We recently identified that the orbitofrontal cortex plays an important role in signalling contextual switches in human tactile reversal learning. However, the neurobiological mechanisms that underlie such signals during behavioural flexibility remain unclear. To investigate this, we designed a probabilistic tactile ‘Go/No-go’ reversal learning task and simultaneously recorded brain-wide EEG responses from healthy participants. Participants first learned to associate specific braille stimuli with ‘Go’ or ‘No-go’ responses and re-leant the task following a rule-switch. To understand the distinct strategies subjects employ to learn the task, we constructed a Bayesian evidence accumulation model. The rule-switch caused an increase in the ‘win-stay’ strategy following ‘false’ rewards (rewards from incorrect decisions), whereas the ‘lose-shift’ strategy following ‘true’ errors’ decreased. Additionally, participants used ‘win-stay’ strategy more following immediately preceding ‘true’ rewards compared to accumulated ‘true’ rewards in recent history. We used event-related potential (ERP) analysis of the EEG responses to investigate rule-specific neural responses and employed imaginary phase locking value (IPLV) analysis to assess context-dependent connectivity changes across brain areas. We observed significant differences in the ERP of ‘false’ and ‘true’ rewards preceding a ‘win-stay’ decision, evidencing context-tracking mechanisms in frontal areas. IPLV analysis revealed dissociated changes in reward- and error-trial-related theta synchronisation between frontal (the dorsolateral prefrontal cortex), parietal, and lateral sensory areas following a rule-switch. These results suggest that frontal areas differentially use theta signals following reward and error feedback to broadcast contextual switches to downstream brain areas.