The size of motor units has been measured in adult rat muscles reinnervated by active and inactive motor axons. The results suggest that active nerve terminals have a competitive advantage over inactive terminals during neuromuscular synapse elimination. The experiments were done using the fourth deep lumbrical muscle in the rat hind foot, which receives its motor innervation from the lateral plantar nerve (l.p.n.) and the sural nerve (s.n.). The muscles were denervated by a nerve crush close to the muscle. Five to ten days later, nerve impulse conduction in the l.p.n. was blocked for 1-2 weeks by chronic superfusion of the nerve with tetrodotoxin (Betz, Caldwell & Ribchester, 1980 b). After 2 weeks of l.p.n. block, the isometric tetanic tension of s.n. motor units increased about two-fold, compared with contralateral control muscles. This was due to an increase in the number of muscle fibres innervated by s.n. motor axons. Intracellular recordings showed that more fibres were innervated by the s.n. than in normal muscles. In some animals, the blocked l.p.n. was cut 1-2 weeks later. The l.p.n. terminals were allowed to degenerate for 1-2 days. There were more s.n. terminals in zinc iodide-osmium stained preparations of these muscles than in normal muscles. Calculation of tetanic tension overlap between l.p.n. and s.n. motor units, and the amount of mono-neuronal innervation seen in intracellular recordings suggested that a larger fraction of the muscles was innervated only by s.n. motor nerve terminals than in controls. This fraction increased with time, ultimately reaching about 14% of the muscle per s.n. motor unit. The expansion of the s.n. motor units appeared to take place by terminal and preterminal sprouting of motor axons. The l.p.n.-evoked tetanic tension decreased in parallel with the increase in the s.n. tetanic tension. The decrease in the l.p.n. twitch tension did not parallel the increase in the s.n. twitch tension. At least part of this discrepancy was due to repetitive firing of the regenerated, inactive l.p.n. terminals when the nerve was stimulated electrically. The results support the notion that modifications in connectivity between pre- and post-synaptic cells can come about by growth or withdrawal of terminals, as a result of differences in the level of activity in the presynaptic cells (Hebb, 1949).