Removal of the caudal half of a goldfish optic tectum induces optic fibres from the entire contralateral retina to terminate retinotopically within the remaining half. This compression has been viewed by some as the result of competition between the fibres and by others as a consequence of changes, induced by the surgery, in tectal labels guiding fibres to terminal sites. To distinguish between these possibilities, the time-course of compression has been measured by electrophysiological mapping of the visual projection. In some fish, fibres terminating in the rostral half-tectum remained intact when the caudal half was removed. In others, the optic nerve was cut at the time of tectal surgery: even after its regeneration into a half-tectum, optic terminals were first detected in the regions they normally occupy. The subsequent reorganization was gradual and retinotopic order was maintained. However, it was slower where some fibres had never been cut. In a third series the nerve was cut 18 days before the tectal halving to reveal any dependence of compression on progressive changes in the halved tectum; but its time-course from nerve section was found to be independent of the time within the regeneration period at which the tectum was halved. In a fourth series the nerve was cut at the time of tectal halving and then cut again after 85--97 days when compression was complete to reveal any permanent change in the halved tectum. No change was evident: the previous compression did not preclude subsequent regeneration of an uncompressed projection and its gradual recompression as before. In a fifth series, repeated crushing of fibres normally ending in the missing caudal tectum temporarily prevented compression among the remainder, while crushing of fibres destined for rostral tectum caused transposition of the remaining projection to the rostral half. Surgically induced changes in the labels which are thought to guide growing fibres to their normal tectal regions do not account for these results. Indeed, this guidance persists unchanged for fibres regenerating a second time after compression. Since compression is delayed while certain fibres are withheld, it appears instead to be the direct result of competition between the fibres. The maintenance of retinotopic order in compression, despite unchanged tectal guidance, may require selective interactions between fibres from different retinal regions which could contribute to the refinement of the normal visual projection.