We have used toxin-mediated ablation to study some aspects of visual system development in Drosophila melanogaster. To devise a method that permits the conditional expression of a cellular toxin, we introduced an amber mutation into the diphtheria toxin-A-chain gene. In transgenic animals, this toxin gene can be activated by providing the gene for an amber suppressor tRNA. By coupling this toxin gene to the photoreceptor cell-specific promoter of the chaoptic gene, photoreceptor cells could be specifically ablated during development. Photoreceptor cell-specific markers normally activated during pupal development failed to appear after midpupation. Photoreceptor cells were absent from the retinas of adult flies at eclosion. We have assessed the consequences of photoreceptor cell ablation for eye and optic lobe development. We suggest that the larval photoreceptor nerve is not essential, in the late larval stages, for retinula photoreceptor cell axons to achieve their proper projection pattern in the brain. Moreover, while retinula photoreceptor innervation is initially required for the development of normal optic ganglia, the ablation of these cells in midpupation has no discernible effect. This approach to cell-specific ablation should be generally applicable to the study of cellular functions in development and behavior.