The effect of muscle length on the probability of neuromuscular transfer for whole frog muscle was studied in vivo at different stimulus frequencies. First the relationship between the steady state amplitude of the compound muscle action potential and stimulus frequency was measured at standard muscle lenght, l1/2, which is the mean of maximal and minimal muscle length in situ. The result shows that the collection of nerve muscle junctions acts as a "low pass filter" for neuromuscular spike transmission with a 50% "pass through" at about 1 Hz for the standard length. At a given stimulus frequency in the range of 0.125 to 8 Hz, a stepwise increase of muscle length from l1/2, within the physiological range, was found to cause an increase of the compound muscle action potential amplitude with an initial overshoot, while a decrease of length caused a decrease of the amplitude with an initial undershoot. The relation between the amplitude of the initial response and muscle length depends on stimulus frequency, and on the degree of curarization. Its shape is sigmoidal at about 50% spike transmission. The results suggest that the maximal physiological change of muscle length can change the difference between the average endplate potential peak amplitude and the average threshold potential up to a value, which is equal to two times the standard deviation of the endplate peak potential distribution of the junctions. For the steady state response to changes in muscle length this effect is smaller. We suggest that the observed effect of muscle length on neuromuscular transmission plays an important role in the control of whole muscle contraction.