1. As a human subject slowly increases the amount of force exerted by a muscle, the discharge rates of low-threshold motor units saturate at a rather low level, whereas higher-threshold units continue to be recruited and undergo increases in their discharge rates. The presently known intrinsic properties of motor units do not produce this "rate limiting." 2. Using computer simulations of a model motoneuron pool, we tested the hypothesis that rate limiting can be accounted for on the basis of the known distributions of synaptic input from different sources. The properties of the simulated motor units and their synaptic inputs were based as closely as possible on the available experimental data. A variety of simulated synaptic input organizations were applied to the pool, and the resulting outputs were compared with the data on rate limiting in human subjects. 3. We found that the data on rate limiting in human subjects greatly constrained the possible organizations of characterized synaptic input systems. Only when the synaptic organization included a gradual "crossover" between two specific types of input systems could the human data be accurately reproduced. Low input/output levels relied on a system organized like the monosynaptic Ia input, which produces greater effective synaptic currents in low- than in high-threshold motor units. Above a sharply defined crossover level, all further increases in output were produced by a system organized like the oligosynaptic rubrospinal input, which generates the opposite pattern.