Previous anatomical studies suggest that androgen regulates synapse elimination in the androgen-sensitive levator ani(LA) muscle of the rat. Androgen treatment beginning on postnatal day 7 (P7) prevents some of the normal loss of multiaxonal innervation in this muscle. The present study used physiological techniques to measure the number and size of LA motor units during the synapse elimination period in muscles from normals, and castrates treated with either testosterone propionate or oil. The number of increments in LA twitch tension as nerve stimulation intensity increased, a measure of the number of motor units, was the same at the end (P28) of synapse elimination as near the beginning (P7) of this process. This result indicates that motoneuronal cell death does not contribute to synapse elimination in the LA. Moreover, androgen during this period did not influence the number of LA motor units. In contrast, between P7 and P28, there was a dramatic decline in the size of LA motor units, as indicated by a decrease in the percentage of twitch or tetanus tension of individual motor units relative to the maximal twitch or tetanus tension of the whole muscle. In addition, androgen treatment of castrated males during this period prevented some of the normal decline in the size of LA motor units. Estimates of the number of inputs per LA muscle fiber derived from the number of LA motor units and their average size indicate that androgen maintains polyneuronal innervation in the LA muscle. This finding supports previous anatomical studies suggesting that androgen can prevent synapse elimination in this muscle. The strength of LA synapses was also examined by measuring the tetanus: twitch ratio of individual motor units and by measuring the safety margin of LA synapses. Both measurements indicated that the average strength of LA synapses increases during synapse elimination. Moreover, androgen appeared to spare synapses from elimination without increasing their strength, since androgen-treated muscles generally had larger motor units but the same mean tetanus:twitch ratio and safety margins as untreated LA muscles except at P28, when synapses in androgen-treated LA muscles had appreciably lower safety margins than normal. These results suggest that androgen regulates synapse elimination through a mechanism(s) independent of synaptic strength.