According to Sperry's chemoaffinity hypothesis, the projection of a small eye fragment with a reduced amount of optic fibres should be restricted to that position in the optic tectum corresponding to its own specificity. However, previous investigations on different types of quarter-eyes in Xenopus laevis have revealed that their retinal projection was always restricted to the rostral part of the tectum, no matter what the origin of the remaining retinal quadrant. To get an indication of the state of specificity in such eye fragments, we investigated by electrophysiological and histological methods several features of the retinal projections of tempo-roventral (TV), naso-ventral (NV) and ventral (V) quarter-eyes which referred to their positional identity. Irrespective of their different origins, the projections were always located in the rostral part of the tectum, the size of the innervated tectal area depending for all fragment types on the size of the quarter-eyes, i.e. number of optic fibres. However, quantitative analyses revealed that with increasing eye size the various fragments expand their projections preferentially into those tectal areas that match their original specificity: TV projection is more concentrated in the rostral tectum, NV eyes expand their projections mainly to the caudal tectum, and V eyes enlarge their projections equally into the medial and caudal tectum. In addition, fibre-tracing experiments with cobaltic lysine showed that, according to the different origins of the quarter-eyes, retinal fibres follow the appropriate branch of the optic tract selectively: fibres of NV and V eyes pass mainly through the medial tract, and most fibres of TV eyes innervate the rostral tectum directly from a central position between the two side branches. All these findings suggest that the different types of quarter-eyes retain their original positional identity. Thus, their rostrally located retinotectal projections are not in register with their retinal specificity. We conclude that in X. laevis local positional markers in the tectum, if present at all, do not influence the development of the retinotectal projection. Instead we suggest a concept of self-sorting of the optic fibres, which can account for the partial innervation of the rostral tectum in different types of quarter-eyes.