In Amphibia all primary retinotectal nerve fibres cross at the chiasma. To investigate why this decussation takes place, a series of embryological operations were performed in which one host eye was replaced by an eye from the opposite side of a donor in Xenopus laevis. The visual projections to the optic tecta were mapped electrophysiologically in these animals when adult to reveal to which side of the brain the nerve fibres had become connected. In the majority of cases the contralaterally grafted eyes had developed primary projections to both tecta, in others all the fibres had innervated the contralateral tectum and in 2 animals optic axons were detected only from the ipsilateral side. In 2 other animals optic nerve fibres from both the normal and operated eyes had innervated both tecta. Control experiments showed that section of the optic nerve and stalk without any misalignment of the eye and stalk did not disturb the normal decussation of the fibres. Neither did enucleation in the embryo affect the decussation of the fibres from the remaining eye. It is concluded that no "side specificity" has been shown to exist to determine the side of the brain into which retinotectal fibres develop; rather it seems that complete decussation depends on the developing optic nerve fibres being able to enter a normally aligned optic stalk as they leave the eye.