1. Transmitter release has been studied at normal, myasthenia gravis (m.g.) and myasthenic syndrome (m.s.) affected human end-plates. At normal and diseased end-plates evoked transmitter release is Poisson for a mean quantal content, m less than ten. 2. The relation between log m and log [Ca]o, at normal and m.g. end-plates is linear, with a slope of 3.3-3.4. The value of m at m.g. end-plates is about five times larger than normal, below Ca 0.7 mM (Mg, 2mM). This difference in m is reduced at higher Ca levels. 3. The slope of the relation between log m.e.p.p. frequency and log [K]o is similar at normal and m.g. end-plates. Over its linear portion the relationship has a slope of approximately 6. 4. Fluctuations in the latency of evoked transmitter release were compared at normal and m.g. nerve terminals. At normal end-plates the probability of release reaches a peak about 0.3-0.4 msec after unitary e.p.p.s of the shortest latency and returns to zero about 1.0 msec after the peak. At m.g. end-plates the distribution of latencies shows less uniformity. 5. At m.s. end-plates m is approximately 5 in normal Ringer solution (2 mM-Ca, 1 mM-Mg). The relation between log m and log [Ca]o is linear, with a slope of 1.0-1.5. The K dependence of m.e.p.p. frequency appears reduced at m.s. end-plates. 6. Assuming a co-operative mechanism for transmitter release at normal human motor nerve terminals, the dissociation constant for the Ca complex is about 1.6X10(-3) M and the dissociation constant for the Mg complex is about 1.0X10(-3) M. 7. It is concluded that the presynaptic changes, at m.g. end-plates, are not the primary cause of the defect in nerve muscle transmission. At m.s. end-plates the presynaptic changes are sufficient to account for failure in transmission. Possible mechanisms for the abnormalities in transmitter release are considered.