The corticostriatal projection in rat neostriatal grafts was studied by using the axonal transport of Phaseolus vulgaris-leucoagglutinin. The neostriatal primodia from 15-18-day embryos were used to make a cell suspension which was implanted unilaterally into the rat neostriatum 3-5 days after kainic acid lesion. Two to four months later, regions of the frontal cortex ipsilateral to the grafts were injected iontophoretically with Phaseolus vulgaris-leucoagglutinin. There were many Phaseolus vulgaris-leucoagglutinin labeled cortical fibers in the host neostriatum. Although the density of labeled fibers in the grafts was much lower than that in the surrounding host tissue, some fibers could be seen to enter the grafts and form terminal arborizations. The morphology of labeled fibers in the graft differed from that of corticostriatal fibers from the same injection but distributing in the host neostriatum. The labeled fibers in the host neostriatum arborized in an extended pattern, branching infrequently and making most of their synapses en passant at varicosities along their courses. The labeled fibers in the grafts made more dense arborizations with many short branches that formed clusters of terminals confined to small foci along their courses. The cellular composition and the structure of the neuropil in the neostriatal grafts were similar to that of the neostriatum. As those in the host, labeled corticostriatal terminals in the grafts contained densely packed round vesicles and made asymmetric synapses on dendritic spines, dendritic shafts and somata. A quantitative analysis, however, revealed that the distribution of postsynaptic elements of labeled boutons in the grafts was different from that in the hosts. More than 90% of the labeled cortical terminals in the host neostriatum contacted dendritic spines whereas only 47% of the labeled terminals in the grafts contacted spines, and 50% of them terminated on the dendritic shafts. The present study provides direct anatomic evidence to demonstrate the restoration of the corticostriatal projection in grafts. The difference in the distribution of postsynaptic elements in the grafts and the hosts may represent a response to the decreased innervation density of cortical inputs to the graft tissue, and may contribute to the recovery of corticostriatal responses by increasing the effectiveness of transmission by the fibers that do grow into the graft and form contacts there.