Neuroimaging techniques such as positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) are widely used to identify the cerebral correlates of cognitive tasks. The resting state presents the advantage to serve as a reference in all experiments but is also an ill-defined mental state because it may vary both from one subject to another and within the same subject. The most challenging question concerns the areas whose activity (revealed by PET or fMRI imaging) is greater in rest state than in an active condition. The present work reports the result of a meta-analysis including five previously published studies. The five different tasks involved are the following: attribution of intention, judgement of stimulus pleasantness, discrimination of spatial attributes, judgement of other peoples' belief and perception of gaze. For each study, the general linear model was used to assess statistical difference and a contrast resting state minus other conditions was calculated. The intersection of the five contrasts was used to search for the variation jointly observed across the different experiments. This lead to a reduced number of clusters: one cluster in the lower/anterior part of the cingulate gyrus and four clusters located in the medial/superior frontal gyrus, along the superior frontal sulcus. We discuss the location of these areas with respect to the location of activations induced by different tasks: externally focused attention, memory, general reasoning, theory of mind and self-referential tasks. We observed that medial prefrontal cortex exhibits a lower activity when the subject's attention is focused towards the external world than when the subject has to additionally refer to some internal states. By contrast, this activity is greater during resting state than during both externally directed and internally directed attention. Thus, we hypothesize that during rest, the subject is in a state where he refers only to his own self.