Focal recording from active spots of a neuromuscular junction was used to measure the 'synaptic delay' between terminal axon spike and end-plate current (e.p.c.). Synaptic delay is defined as the time interval between peak of inward current through the presynaptic membrane and commencement of inward current through the postsynaptic membrane. By substituting magnesium for calcium in the medium, and by adjustable electrophoretic application of calcium from the recording electrode, the e.p.c. can be restricted to the small portion of a single junction which is in contact with the microelectrode, and the statistical average amplitude of the e.p.c. can be reduced to less than quantal unit size. Under these conditions, the latency of the unit components of the e.p.c. can be determined and its statistical fluctuations studied. The synaptic delay at a single end-plate spot has a minimum value, at 20 degrees C, of 0$\cdot $4 to 0$\cdot $5 ms and a modal value of about 0$\cdot $75 ms. There is considerable fluctuation of the measured intervals during a series of nerve impulses; over 50% occur within a range of 0.5 ms, the rest being spread out in declining fashion over a further 1 to 4 ms. These latency fluctuations are shown to be a statistical consequence of the quantal process of transmitter release. The contribution of various factors to the minimum synaptic delay are discussed. Terminal conduction time has been effectively eliminated by the method of focal recording. Diffusion of acetylcholine towards the receptors, and its reaction with them must cause delays whose exact values are uncertain, but whose extreme upper limits can be shown to make up only a small part of the observed minimum delay. It is concluded that the synaptic interval arises chiefly from a delay in the release of transmitter after the arrival of the nerve impulse.