Deciding for a course of action requires both an accurate estimation of option values and a right amount of effort invested in deliberation to reach sufficient confidence in the final choice. In a previous study, we have provided evidence, across a series of judgement and choice tasks, for a dissociation between the ventromedial prefrontal cortex (vmPFC), which would represent option values, and the dorsomedial prefrontal cortex (dmPFC), which would represent the duration of deliberation. Here, we first replicate this dissociation and extend it to the case of an instrumental learning task, in which 24 human volunteers (13 women) choose between options associated with probabilistic gains and losses. According to fMRI data recorded during decision-making, vmPFC activity reflects the sum of option values generated by a reinforcement learning model, and dmPFC activity the deliberation time. To further generalize the role of the dmPFC in mobilizing effort, we then analyze fMRI data recorded in the same participants while they prepare to perform motor and cognitive tasks (squeezing a handgrip or making numerical comparisons) to maximize gains or minimize losses. In both cases, dmPFC activity is associated with the output of an effort regulation model, and not with response time. Taken together, these results strengthen a general theory of behavioral control that implicates the vmPFC in the estimation of option values and the dmPFC in the energization of relevant motor and cognitive processes. Significance statement The medial prefrontal cortex (mPFC) is known to represent key variables needed for choosing a course of action. We previously suggested a functional partition of this brain region: the expected values of choice options are signaled by the ventral part (vmPFC) and the effort invested in decision-making by the dorsal part (dmPFC). Here, we generalize this functional partition to various motor and cognitive tasks, using fMRI in healthy volunteers. Results show that vmPFC activity reflects the expected value of options generated by a reinforcement learning model (whether the goal is to maximize reward or avoid punishment), while dmPFC activity reflects the output of an effort regulation model (whether the task is to produce force or to compare digits).