The concept of self has evolved over time to become one of the central pillars of the human mind. Work over the last two decades have began to relate the concept of self to large scale networks of the brain¹ and especially with the default mode network (DMN)². Recent work from our team ³has further suggested that the neural representation of self (NRS) is confined to two nodes at the very centre of the DMN. A node in the dorsal midline frontal cortex is likely the executive part of the NRS while a node on the mid-parietal cortex is likely the depository of the memory store of the NRS. The NRS may well be the pinacle of evolution in humans, but it also creates a conflict. For a number of fundamental evolutionary reasons ^{3, 4} the NRS must retain its individuality intact while the world outside changes, often in dangerous ways, demanding that these changes are also reflected in the internal representation of the world. In this context safe learning emerged as a way of achieving experience-guided change of the internal representation of the world that makes only the minimal necessary changes to the NRS. It seems that evolution has resolved the conflict through seggregation of learning in time. Learning when practically no change in the NRS is incolved is allowed in real time, while learning that also demands changes of the NRS, is allowed in the form of safe learning only during special times, mainly during sleep. Detailed studies of the changes during light sleep have revealed how this safe learning (memory consolidation) is prepared during spindles ⁵: first the environment is checked and if judged as safe then the alerting system is actively inhibited so that influences from the environment are blocked; then and only then, memories that are temporarily stored in mid-temporal brain areas are highlighted for spindle-mediated transfer to widespread brain areas. The changes of the NRS seem to be controlled by the rostral, executive, part of the NRS as the actual NRS-related memories are transferred to its caudal part ³. In the framework described above neurofeedback is seen as a safe yet incomplete process to reverse aberrant behaviour to physiological norms. The process is safe because it relies and exploits the natural ways of safe learning described above and for this reason it is incomplete: whatever is achieved during neurofeedback corresponds to directed experience related to the NRS that will be developed further during the day, but only consolidated during sleep, because only then the NRS is open for change. In the context of helping people stop smoking in the SmokeFreeBrain project, the influence of the object of addiction that overrides the normal controlling influence by the NRS is confronted by neurofeedback in two stages; in the first stage a small number of sessions are used to bring the NSR closer to normal physiological patterns. In the second stage the well-established alpha-theta protocol is used to address the addiction directly. Early results from the cases already completed show that both neurofeedback stages have a positive influence, with progress in the first stage beeing a pre-requisit for successful outcome of the combined sessions.