After unilateral ablation of the optic tectum in the frog (Rana pipiens), retinal ganglion cell axons enter the lateral thalamic neuropil in large numbers. This area is normally a target of the tectal efferent projection but is not innervated directly from the retina in normal frogs nor in frogs undergoing optic nerve regeneration in the presence of an intact tectum. The ability of retinal axons to form synaptic contacts in this nonstandard target, previously suspected only from light microscope studies, has been ultrastructurally verified in the present investigation. Retinal axon terminals were selectively labeled for light and electon microscope study by introducing horseradish peroxidase (HRP) into the optic nerve 73-413 days after unilateral ablation of the contralateral optic tectum. In some of the frogs, the optic nerve had also been crushed to test the ability of retinal axons regenerating over a long distance to form this connection. The HRP-labeled retinal axon terminals had the same untrastructural morphology whether located in the lateral thalamic neuropil or in the correct regions of projection, e.g., the lateral geniculate complex. They contained clear, spherical synaptic vesicles and made Gray type I synapses on the unlabeled postsynaptic dendrites. The magnitude of the projection was disproportionately greater in animals having complete or nearly complete tectal ablation than in a specimen in which the lesion was significantly incomplete. An aberrant projection was also observed in the nucleus isthmi in some of the specimens. These findings have significance for chemoaffinity theories of the specification of synaptic connections in that the ability of retinal axons to synapse in nonstandard targets in this experimental context may be considered evidence for the expression of appropriate cell-surface recognition-molecules by the abnormally targeted postsynaptic neurons. The likelihood that the expression of these postsynaptic labels is normally repressed transynaptically by molecular signals from the intact tectal input is discussed.