At the time of synaptogenesis typically 50% of the neurons die. The biological role of this is still unclear, but there is evidence in the visual system that many neurons projecting to topographically inappropriate parts of their target are eliminated to improve the accuracy of the mapping. The signaling that determines neuronal survival involves electrical activity and trophic factors. Based on these observations, we have elaborated a computational model for the self-organization of a two-layered neural network. We observe changes in the topographical organization between the two layers. In layer 1, a traveling wave of electrical activity is used as input. Activity transmission to layer 2 can generate, according to a Hebbian rule, a retrograde death signal that is compensated by a trophic survival signal generated by the target cells. Approximately 50% of the neurons die, and we observe refinement in the topography between the two layers. In alternative versions of the model, we show that an equivalent reorganization can occur through Hebbian synaptic modification alone, but with less precision and efficiency. When the two mechanisms are combined, synaptic modification provides no further improvement over that produced by neuronal death alone. This computational study supports the hypothesis that neuronal death during development can play a role in the refinement of topographical projections in the nervous system.