The part played by synaptic plasticity in natural learning can be reassessed owing to the discovery of dormant neurons in the cortical layer-2 of large-brained mammals. Here, the recruitment of cortical columns that contain these quiescent cells is argued to underlie the lifelong building of long-term memories, albeit only during periods when cortical regions are segregated, while perception, cognition and motor skills are at their lowest level. As records of online experience are replayed by the hippocampus, about half of every sleep cycle would harbour the relevant neurites outgrowth between in-use and free cortical columns. This translation of labile traces into steady imprints is simulated on computer, in an attempt to link experimental findings to specific functions and possibly generate predictions. Rapid eyes movement observed at the end of every sleep cycle is notably interpreted as a side effect of the necessary priming of peripheral effectors whereby cortical replay can induce activation in the hippocampus and thus allow new connections between paths that are thus concurrently activated. Architectural and functional data highlight a topological matching between key components of the underlying model and those of thalamocortical structures. Several factors are considered for the experience-driven development of these structures, including slow-wave activity whereby the guided growth of axons may occur from layer-2 of their host column towards the layer-4 of free columns, possibly steered in parallel by distinct calcium waves. In addition to this main prediction, this theory supports the central role of sleep in the development of autistic symptoms. Rooted in an epigenetic downregulation of the MAOA brain enzyme, the resulting offline over-activity would entail aberrant branching between cortical columns. Whereas most features of the condition have already been interpreted through this systemic approach, the evolution of autism prevalence across four human generations is derived here from the masking effects of genetic variants and X chromosome inactivation, while a host of risk factors is liable to promote MAOA. Accordingly, pregnancy should be prevented from excess MAOA activity by being subjected to prior blood testing, which may lead to the individualized prescription of MAOA inhibitors.