The visual system of lower vertebrates has served as an important testing ground for the mechanisms that generate topographic neuronal connections. During both the outgrowth and the regeneration of the optic nerve, a smoothly ordered map of the retina is formed on its major target, the optic tectum (the retinotectal projection). Experiments performed on this projection have offered support for a variety of mechanisms, including the matching of positional cues in the retina and tectum, the guidance of nerve fibers by interactions between fibers, competition for synaptic space, and the refinement of connections based on neuronal activity. Unfortunately, individual experiments that support any one of these mechanisms have been taken at times as evidence against the involvement of any other mechanism; for example, experiments demonstrating the importance of positional cues have been thought mistakenly to indicate that activity-based interactions are unimportant. Computer simulations, in which multiple, somewhat opposed, mechanisms are allowed to operate in concert demonstrate that such a hybrid model is able to generate a full range of experimental results. More importantly, the elimination of any one of the mechanisms renders the model unable to fit entire classes of findings. Thus, the patterning of the retinotectal projection is best viewed as a process in which the optic nerve terminals attempt to satisfy multiple constraints in selecting their target sites.