Decisions are often made in the face of uncertainty and in the absence of immediate feedback. In these scenarios a metacognitive sense of confidence in having made a correct choice can be used to guide future behaviour. Recent computational models propose that our sense of confidence reflects an estimate of the probability that a choice is correct. However it has proven difficult to experimentally separate decision confidence from its component parts, such as our certainty about perceptual evidence or requirements for explicit reports. We devised a task to dissociate these quantities and isolate a distinct encoding of decision confidence in the perigenual anterior cingulate cortex (pgACC) of the human brain. We show that fMRI activity in this area not only tracks expected performance on a task, but also is related to both within- and between-subject variation in a subjective sense of confidence. In contrast, brain areas previously proposed to encode decision confidence instead tracked sensory reliability (posterior parietal cortex and ventral striatum) or criterion distance (presupplementary motor area). In my talk I will expand on these findings in light of broader theories of the contribution of the human frontal lobe to metacognition. I will argue that metacognition research holds particular promise for shedding light on disorders of mental health, as it seeks to develop models of how we form beliefs about ourselves at multiple levels of abstraction.