Many everyday life predictions rely on the experience and memory of event frequencies, i.e., natural samplings. We used functional magnetic resonance imaging (fMRI) to investigate the neural substrates of prediction under varying uncertainty based on a natural sampling approach. The study focused particularly on a comparison with other types of externally attributed uncertainty, such as guessing, and on the frontomedian cortex, which is known to be engaged in many types of decisions under uncertainty. On the basis of preceding stimulus cues, participants predicted events that occurred with probabilities ranging from p = 0.6 to p = 1.0. In contrast to certain predictions in a control task, predictions under uncertainty elicited activations within a posterior frontomedian area (mesial BA 8) and within a set of subcortical areas which are known to subserve dopaminergic modulations. The parametric analysis revealed that activation within the mesial BA 8 significantly increased with increasing uncertainty. A comparison with other types of uncertainty indicates that frontomedian correlates of frequency-based prediction appear to be comparable with those induced in long-term stimulus-response adaptation processes such as hypothesis testing, in contrast to those engaged in short-term error processing such as guessing.