The retinotectal projection is a premier model system for the investigation of molecular mechanisms that underlie axon pathfinding and map formation. Other important features, such as the laminar targeting of retinal axons, the control of axon fasciculation and the intrinsic organization of the tectal neuropil, have been less accessible to investigation. In order to visualize these processes in vivo, we generated a transgenic zebrafish line expressing membrane-targeted GFP under control of the brn3c promoter/enhancer. The GFP reporter labels a distinct subset of retinal ganglion cells (RGCs), which project mainly into one of the four retinorecipient layers of the tectum and into a small subset of the extratectal arborization fields. In this transgenic line, we carried out an ENU-mutagenesis screen by scoring live zebrafish larvae for anatomical phenotypes. Thirteen recessive mutations in 12 genes were discovered. In one mutant, ddl, the majority of RGCs fail to differentiate. Three of the mutations, vrt, late and tard, delay the orderly ingrowth of retinal axons into the tectum. Two alleles of drg disrupt the layer-specific targeting of retinal axons. Three genes, fuzz, beyo and brek, are required for confinement of the tectal neuropil. Fasciculation within the optic tract and adhesion within the tectal neuropil are regulated by vrt, coma, bluk, clew and blin. The mutated genes are predicted to encode molecules essential for building the intricate neural architecture of the visual system.