Neural assemblies are assumed to become organized and to operate within the cerebral cortex, and so must be constrained by the cytological and physiological properties of this laminated structure. A hypothesis of such assemblies is presented, based on important details of neuronal architecture and physiology in different cortical laminae. Laminae II, III and VI, which are the origin and termination of most cortico-cortical projections, are regarded as the site of storage of most of the information encoded by assemblies - a neuronal 'library'. Laminae II and III are the most sensitive coincidence detectors, and therefore probably initiate the process of assembly formation. However, these three laminae have very low levels of spontaneous activity in the waking state, and so active cell assemblies cannot base their functioning on these laminae alone. Lamina V pyramidal cells have a much higher level of spontaneous activity. Thus, indirect pathways between 'library' cells, via lamina V pyramidal cells, are likely to be more secure than direct ones. It is proposed that direct links between 'library' cells become stabilized by Hebbian strengthening, once the recipient 'library' cell has been 'primed' by neural activity transmitted indirectly via lamina V neurones. Thus lamina V neurones could catalyse the process of assembly formation. Given this proposal, lamina V cells, in their interaction with 'library' cells, would code information in terms of precisely timed individual impulses, but would employ a code based on slower frequency changes in their descending influences upon neural centres in the brainstem and spinal cord. Predictions for single unit and electrographic experiments are discussed.