The dorsolateral quadrant of the lateral septal nucleus receives projections from both the ipsilateral and the contralateral fimbria. In the adult rat the effect of fimbrial lesions on synapse formation has been studied by a quantitative electron microscopic analysis of the various types of synapses present, using electron-dense degeneration to identify fimbrial fibre terminals. In this area, the fimbrial axons from both sides together account for about 30% of the total number of synapses and they terminate mainly on dendritic spines. The ipsilateral fimbria forms twice as many synapses as the contralateral fimbria. When one fimbria is cut and time left for the degeneration to be removed, the numbers of synapses are restored to normal levels and the remaining fimbria acquires, on both sides of the septum, a number of synapses equal to the sum of the two individual fimbria, This suggests that the axons of the surviving fimbria have completely reinnervated the denervated postsynaptic sites formerly occupied by the cut fimbria of the other side, effectively excluding non-fimbrial axon terminals, even though the latter constitute the majority (70%) of the synaptic terminals in the region. When both fimbria are cut the numbers of synapses are once again restored to normal levels. However, since there are now no fimbrial axons left, the denervated fimbrial postsynaptic sites must this time have been reinnervated by non-fimbrial axons. Reinnervation by non-fimbrial axons is numerically equally effective in reclaiming the denervated sites, although when compared to the reinnervation by fimbrial axons, the removal of degenerating terminals is somewhat slower, and among the reinnervating terminals there is a much higher incidence of axon terminals making more than one synaptic contact in the plane of section. Thus, fimbrial axons, when present, have the ability to exclude the reinnervation of denervated fimbrial sites by non-fimbrial axons, despite the fact that the latter are both more numerous and also clearly capable of reinnervating those sites when no fimbrial axons are present. Two possible mechanisms are discussed: a spatial preference based on the geometrical arrangements in the neuropil, and a temporal preference based on the relative rates of response of the fimbrial vs the non-fimbrial axons.