In normal goldfish, lesions of various size were made in nasal or temporal retina immediately prior to retinal labeling with tritiated proline. The resulting gaps in retinal innervation of tectum indicated that the projection is retinotopographically ordered to a precision of about 50 micrometer. Similarly, acute tectal incisions transecting the optic pathways were combined with immediate retinal labeling. The resulting tectal denervation confirmed that most fibers follow highly ordered paths through the stratum opticum of tectum; but a few fibers were found to follow unusual paths to their appropriate tectal positions. In other fish, the optic nerve was crushed. At various times afterwards, retinotopography and pathway order were similarly analyzed by making retinal lesions or tectal incisions just prior to labeling. For up to 40 days after crush, the projection lacked any refined retinotopic order. Only a gross topography could be demonstrated. Over several months, retinotopography gradually improved eventually approaching that of normals. Correlated with this was an initial stereotypic growth through the pathways of the stratum opticum followed by a long period of highly anomalous growth through the innervation layer. Evidently, many regenerated fibers grew in through inappropriate routes to the wrong region of tectum but subsequently arrived at their appropriate locus by circuitous routes within the innervation layer.