1. End-plate potentials (e.p.p.s) and end-plate currents (e.p.c.s) were recorded intracellularly from muscle fibres of frog and mouse at various levels of curarization to determine the relation between the potential change and the underlying synaptic conductance change over a wide range of e.p.p. amplitudes. 2. In frog muscle fibres the e.p.p.-e.p.c. relation was linear for e.p.p. amplitudes up to about 10 mV, beyond which the rate of increase of e.p.p. amplitude became progressively smaller as the e.p.c. amplitude increased. The equation proposed by Martin (1955) to correct for this non-linearity consistently over-corrected the e.p.p. amplitudes. 3. When synaptic potentials and currents of long duration were produced by ionophoresis of ACh onto the end-plate, the voltage-current relation showed greater non-linearity than with nerve-evoked responses, and correction of the synaptic potential amplitudes resulted in a linear relation. 4. The relation between e.p.p. and e.p.c. amplitudes in mouse muscle showed a greater non-linearity than in frog muscle and over-correction by the equation was correspondingly smaller. Theoretical voltage-current relations were calculated for various membrane models and compared with the relations observed experimentally. The results from mouse muscle agreed with those expected for a point synaptic contact on an infinite cable; those from from muscle were consistent with simple resistive-capacitative model with no cable extending from the synaptic region. 6. The applicability to the experimental results of several correction factors for non-linear summation is discussed.