1. The acetylcholine-sensitive ionic channels at the neuromuscular junction were studied in voltage-clamped single muscle fibres from a monolayer preparation of the cutaneous pectoris muscle from Rana pipiens. The experimental observations were of three types: (a) reversal potential as a function of external Na and Ca concentrations, (b) the single channel conductance (gamma) from noise analysis as a function of these same concentrations, and (c) gamma as a function of membrane potential. 2. The reversal potential in normal Na Ringer was -3.8 +/- 0.5 mV (+/- S.E. of mean, n = 22) and decreased approximately linearly as the logarithm of the outside Na activity as this activity decreased to 10% of normal. 3. The single channel conductance in normal Na Ringer was 27.5 +/- 0.7 pS (n = 28) and reached a limiting value close to 10 pS as Na was replaced with sucrose. 4. Increasing [Ca]o from 2 to 10 mM made the reversal potential more positive and decreased the single channel conductance. Mg caused similar effects. 5. Various theories that have been used to describe the mechanism of ion permeation through e.p.c. channels were tested. Constant field theory (eqns. (3), (4) and (5)), a modified Takeuchi approach (eqn. (6)), and a single barrier theory (eqns. (8), (9) and (10)) could not account for all of the experimental observations. 6. In particular, constant field theory, with no surface charge density, could account for the following: (a) the reversal potential measurements for solutions containing 2 mM-Ca (with PK/PNa = 1.2 and PCa/PNa = 1.02), (b) the single channel conductance values for solutions containing 2 mM-Ca and Na concentrations down to 20% of normal, (c) that gamma has little voltage dependence. 7. However, constant field theory, with no assumed surface charge density, could not account for the following: (a) the reversal potential observed for Ringer containing 80 mM-Ca, (b) the gamma values observed for very low Na concentrations, (c) the observation that increasing Ca from 2 to 10 mM in a solution containing 75% normal Na results in a decrease in gamma. 8. From the failure of the Takeuchi approach (eqn. (6)), it is argued that ion interactions must occur at e.p.c. channels because ion flux independence is the only asumption in the derivation of eqn. (6) without experimental verification. 9. The ion interactions at e.p.c. channels probably include both surface charge effects and competition for a binding site.