In mammals, synaptic connections formed between axons o] retinal ganglion cell and neurons of the lateral geniculate nucleus are complex and yet strikingly precise. This poses a formidable problem for the development o f these connections. Research performed in a number of laboratories during recent years has provided glimpses into many of the interacting mechanisms that work to control this development. Because so much of the relevant research has been performed on cats, this brief review will concentrate on the normal and abnormal development of the retinogeniculate pathway in this species. The reader can find a more thorough discussion of the mammalian visual system and its development in several recent reviews 1~. The evidence discussed below is organized around the hypothesis that development of the cat's retinogeniculate pathway is largely (but not entirely) controlled by competitive interactions among developing axons. The competitive development can at least conceptually be divided into three partially overlapping and related processes. First is the overproduction of ganglion cells and their axons followed by the loss of those that fail to make appropriate connections in the lateral geniculate nucleus; second is the formation of clear lamination in the lateral geniculate nucleus during the same period that retinogeniculate axons develop extensive arbors within their appropriate laminae; and third is the shaping and pruning of arbors as they continue to develop. The first two processes occur mostly prenatally and the third extends several months into postnatal life, although different axonal classes seem to engage in these processes during somewhat different prenatal and postnatal periods. However, the actual evidence presently available makes only a circumstantial if at times compelling case that the overproduction and subsequent death of excess ganglion cells, the development o f lamination, and the pruning of retinogeniculate axon arbors are consequences of competition.