Embryonic development of nerve patterns has fascinated and frustrated neurobiologists for decades. The orderly connections within the central nervous system and between the central nervous system and peripheral end organs are thought to be formed by a number of different means ranging from selective cell death, through timing and mechanical guidance, to chemospecificity. A system that has gained great popularity for the study of topographic connections between sets of neurons is the retinotectal system of lower vertebrates (amphibia and fish). Although this system has hardly been cooperative in yielding conclusive evidence (Horder & Martin 1978, Hunt & Jacobson 1974), its great popularity has led to a mass of experimental data, which, by sheer bulk and ever more refined experimental design, has begun to crystallize into strong support and strong challenge to various proposed ideas for the assembly of connections between the eye and brain.