A problem with the use of muscle grafting as a therapeutic procedure is to produce a graft functionally adequate to replace a muscle of complex architecture, such as a sphincter muscle. We thought it might be possible to use dead cadaver muscles, repopulated by the patient's own muscle precursor cells (mpc), to reconstruct muscles whose anatomy would be imposed by the framework of dead muscle and whose genetic constitution would be determined by the mpc. Here we show, in the mouse, that an extensor digitorum longus (EDL) muscle, killed by repeated freezing and thawing, repopulated with mpc and grafted into a nu/nu or tolerant AKR host mouse, is capable of supporting muscle formation. By using the allotypic isoenzyme forms of glucose-6-phosphate isomerase as markers, we have shown that the newly regenerated muscle in such grafts is derived mainly from the implanted mpc, but also to some extent from the host mouse's own mpc. By 50–70 days after grafting, new muscle fibres were found to constitute up to 70% of the graft. Many fibres had assumed diameters in the normal range for mouse muscle, often having peripherally placed nuclei. These findings raise the possibility of the therapeutic use of such grafts. To our surprise, dead EDL muscle grafts into which no mpc had been implanted were also the site of good muscle regeneration. New-formed muscle in these grafts was shown to be derived entirely from mpc which must have migrated into the graft from the host. Investigation of the mechanisms underlying this phenomenon should further our knowledge of factors which regulate the proliferation and movement of dormant mpc in adult animals.