The medial orbitofrontal cortex (mOFC)/ventromedial prefrontal cortex (vmPFC) has been proposed to signal the expected value of rewards when learning stimuli-rewards associations. Yet, it is still unclear whether identical or distinct orbitofrontal cortex regions encode expected rewards and punishments at the time of the cue during appetitive and aversive classical conditioning. Moreover, it is unknown whether anticipation of different types of positive and negative reinforcers differentially influence specific orbitofrontal cortex regions. To answer these questions, this study investigated whether the human mOFC/vmPFC region encodes a general unsigned anticipatory value signal for different types of rewards and punishments (responding in a positive fashion in anticipation of both appetitive and aversive events) or a signed expected value signal (responding positively in anticipation of rewards and negatively in anticipation of punishments) when learning cue-outcomes associations. Using a model-based fMRI approach implementing a reinforcement learning model to compute the expected values of two types of rewards (pleasant juice, monetary gain) and two types of punishments (aversive juice, aversive picture), we found that mOFC/vmPFC activity correlated positively with the expected value of the cues, in anticipation of both rewards and punishments. This finding indicates that the mOFC/vmPFC encodes a general unsigned anticipatory value signal, regardless of reinforcers valence (positive/negative) and types (gustatory, visual).