The accuracy of reinnervation in peripheral nerves following second degree injuries, which do not disrupt the longitudinal continuity of the endoneurial sheaths, has been studied in rats. The sciatic nerve or lumbar spinal nerves (that is the extraspinal nerves before their fusion in the sciatic plexus) were crushed with fine watchmakers' forceps in neonatal and adult rats. In addition, the lumbar spinal nerves were frozen in a group of 5 adult rats. After allowing reinnervation to occur for 5 to 9 weeks, the motoneurons whose axons ran in the plantar nerve were labelled retrogradely with horseradish peroxidase. Their positions in the grey matter of the lumbar spinal cord were recorded and compared with those labelled from the contralateral unoperated plantar nerve. Very few errors of projection occurred after a crush lesion of the adult sciatic nerve but all the other lesions produced significant numbers of errors. The order, starting with the preparations with fewest errors was as follows (numbers in brackets = percentage of neurons misplaced): sciatic crush in adult (3%), sciatic crush in neonate (23%), spinal nerve freeze in adult (23%), spinal nerve crush in adult (35%), and spinal nerve crush in neonate (72%). It seems that a significant number of axonal growth cones cross endoneurial sheaths after crush or cryoinjuries. Explanations for the difference in observed reinnervation accuracy between young and old rats and between lesions in peripheral nerves and spinal nerves are discussed. The first is that axons in peripheral nerves in older rats have a less penetrable endoneurial membrane encasing them. The second is that the amount of misrouting is the same at all lesion sites but is much less easily detectable after sciatic lesions than spinal nerve lesions. This is because axons are organized in a 'musculotopic' manner in peripheral nerves and exchange of axon positions will occur largely between axons destined for the same peripheral target. In contrast, exchange of positions of axons in spinal nerves will lead to more overt errors because at this site axons destined for particular muscles do not lie side by side but are intermingled with axons innervating other peripheral targets.